Get Started with Imaris 10.1 Yale - 3D-4D Microscopy Analysis

Transcript

• 00:00OK. Obviously we're going to,
• 00:01we're going to record the whole
• 00:03session is being recorded.
• 00:04So you'll get a a copy of the
• 00:06recording when we're all done as well.
• 00:08But right now the the
• 00:09microscopes will be muted.
• 00:11So if there is a question we'll
• 00:13have questions towards the end.
• 00:14You can type them in at any time.
• 00:15I'll try to go through the the questions
• 00:18at the end of the at the end of the
• 00:20session and we'll try to address what
• 00:21we can with the questions there.
• 00:23If I can try to unmute some of you
• 00:25to ask a real question, I can,
• 00:26I can try to do that as well.
• 00:28Or at the very end like.
• 00:30And I said I'll,
• 00:31I'm sharing my e-mail address
• 00:33where you can contact the support
• 00:35team and we can answer your your
• 00:37questions one-on-one directly
• 00:38with it with your own private
• 00:41one-on-one session moving forward.
• 00:42OK.
• 00:42So why don't we just go
• 00:44ahead and get started?
• 00:49So thank you for coming.
• 00:52This is a webinar session to get
• 00:54started with the Morris using the
• 00:56latest version of our software
• 00:58which is currently at version 10.1.
• 01:00My goal here today, my name is,
• 01:03my name is Matthew.
• 01:04My goal here today is to give you a
• 01:07general overall introduction tour Amaris.
• 01:09Now, I'm not going to be able to
• 01:11cover everything in the session today.
• 01:13I'm going to try to talk for
• 01:15about 90 minutes or so to kind of
• 01:17give you an overview of a lot of
• 01:18different features in Amaris.
• 01:19I'm going to walk through some of them.
• 01:21I'm going to kind of just show
• 01:22you the results on others.
• 01:23There's going to be a lot of things
• 01:25that we're going to cover just to
• 01:26kind of give you a general overview
• 01:27of all the different features,
• 01:28all the different types of analysis,
• 01:30most of the most common analysis that
• 01:32you can do with with the software.
• 01:33Now, there's a lot of things
• 01:34I'm not going to cover today.
• 01:36And if there's anything that's in our
• 01:40in our pocketbook of features and
• 01:42modules that we didn't cover today,
• 01:45please reach out to the US support team.
• 01:47The e-mail down here is in the
• 01:48bottom of my signature.
• 01:49This will come to our general support inbox.
• 01:51Depending on where you are
• 01:52and where you're located.
• 01:54I'm your typical primary support support
• 01:56person in the East Coast so I'll
• 01:59probably be the person that'll reach out,
• 02:00but if I'm not around it will go to
• 02:02one of our team members that will
• 02:03assist you and give you the training
• 02:05on your data and and help you with
• 02:06the software as best that we can.
• 02:08Now the format of today is I'm not
• 02:10going to present a whole lot of slides.
• 02:12This is my only slide to just
• 02:14kind of present the title here.
• 02:16What I'm basically going to do is I'm
• 02:18going to walk you through it from start
• 02:20to finish how to use the software.
• 02:22So this is going to be a hands going to
• 02:24be a real time live demo of the software.
• 02:27So I'm going to try to present as best I
• 02:30can as slowly as I can the general features.
• 02:32Not only kind of how to get the data
• 02:35from your raw microscope file into
• 02:36a Mars but then how to analyze,
• 02:39process,
• 02:39create surfaces and export your
• 02:40data as a final product.
• 02:42So it's all going to be kind of hands on.
• 02:44So like I said I'm going to try
• 02:45to go as slow as I can to kind of
• 02:47show you all the features.
• 02:48But again a lot of the stuff
• 02:50is going to be pre created,
• 02:51some of it is going to be brand new.
• 02:54So let me start up a Mars.
• 02:55So when you launch a Mars,
• 02:57the first thing that you're going
• 02:58to see when you launch a Morris
• 03:00is this particular arena view.
• 03:02Now the reason I wanted to kind
• 03:03of start with
• 03:03this is because this is
• 03:05where everything starts.
• 03:05Your data is either in a TIF image,
• 03:08it's a Zeiss image, Leica,
• 03:09Nikon, some sort of microscope
• 03:11format that has some metadata,
• 03:13or it's a TIF series that has no metadata
• 03:15and you acquired it somewhere else,
• 03:16so you converted it from somewhere
• 03:18else and it's some sort of TIF series.
• 03:20We can handle all of these data
• 03:22formats and import them into a Morris.
• 03:24A lot of these file formats are
• 03:27directly are directly compatible
• 03:28with our file converter.
• 03:30So when you open up in Morris and you
• 03:32have a view that looks like this.
• 03:33This Arena view is just a
• 03:35visualization tool to find out and
• 03:37list folders that you want to analyse.
• 03:40Now I already have a bunch of
• 03:42folders listed here on the left
• 03:43here and then inside this folder for
• 03:45example data sharing for courses
• 03:47has a bunch of image files.
• 03:48These image files are just
• 03:50files that I've processed.
• 03:51I'm going to show you what it looks
• 03:53like in an explorer here real quick.
• 03:54It's going to look something like this,
• 03:55where you're going to have a bunch
• 03:57of files in here that are just
• 03:58going to show you the image files.
• 04:00You're not going to see all
• 04:01the different files.
• 04:01So if you had text files in here,
• 04:03PDFs in here,
• 04:04you're not going to see that
• 04:05in the Arena view.
• 04:07You're only going to see the image files.
• 04:09Now most people are going to share a
• 04:11file or folder that has data that has
• 04:14not been converted into the Mars format yet,
• 04:17and so one way to convert that file.
• 04:18There's two ways to convert those files.
• 04:20One way way to convert those
• 04:21files is to share a file folder.
• 04:23Share your folder and you're going to get
• 04:24a a look of some data that looks like this.
• 04:26If you have a a native Mrs.
• 04:29file,
• 04:29it's going to look like this.
• 04:30You can see that there's no little
• 04:32arrow next to the icon here.
• 04:34That means that file has not been converted.
• 04:36That file has already been converted to Mrs.
• 04:38file.
• 04:38And so you can take that file,
• 04:40double click on it,
• 04:41and it'll open up and you'll
• 04:42you'll look at that file.
• 04:45The icon here identifies
• 04:46that that's a 3D data set.
• 04:48If it's a 2D data set,
• 04:49it'll look like a little square.
• 04:51If it's a time lapse,
• 04:52it'll have another icon that'll
• 04:53show you that it's a time lapse.
• 04:54It is. So there's a little bit
• 04:55of information on the file,
• 04:56as well as the size of the file,
• 04:58and the pixel size of the file
• 05:00is listed there in this kind
• 05:01of view that we have right now,
• 05:02which is our detailed view of the file.
• 05:05Now, if it does have a little
• 05:06arrow next to it,
• 05:07that means you need to convert this file
• 05:08before you can go into the next step,
• 05:10which is to open up the
• 05:12file inside of Remars.
• 05:13Now, like I said,
• 05:13a lot of the file formats are
• 05:15directly supported by Mars.
• 05:16Not all of them,
• 05:17but most of the major formats are.
• 05:19So I'm listening a bunch of
• 05:20the major ones here.
• 05:21There's a Nikon file,
• 05:22there's a CZI file,
• 05:24and I think I have a lift file,
• 05:26and I don't have a LIF file,
• 05:27but the LIF files look the
• 05:29same as the CZI files.
• 05:30There's an OIB file that's Metamorphile.
• 05:33There's a lot of different types of
• 05:35file formats that we support Now.
• 05:37The easy solution to convert these
• 05:38files is just to come in here.
• 05:40Right click on the file and it
• 05:42says convert to native format.
• 05:43Now again if it's supported by our
• 05:45file format and you hit convert,
• 05:47you're going to see a little
• 05:48conversion down here.
• 05:49Process depending on the size of
• 05:50the file might be really quick.
• 05:51If it's a really large file
• 05:52it makes it take some time.
• 05:54But now that that file has been
• 05:55converted you see the error disappears.
• 05:56Now you can click on that file and
• 05:58it'll open up that inside of the Morris.
• 06:01Now the other thing I wanted to
• 06:02mention here because it's really
• 06:03important I just spend a little
• 06:04bit of time on this is that some
• 06:06files have multiple components where
• 06:07you save aczi file like a files.
• 06:09Nikon files I think do this as
• 06:11well if you have that option.
• 06:13There's multiple images inside of
• 06:15that file format and so the idea
• 06:18here is if you click on if you share,
• 06:21open up that file like this CZI
• 06:23file has multiple components,
• 06:25so there's actually 3 images
• 06:26in that one CZI file.
• 06:27So if we come over here and look at
• 06:29that CZI file, that CZI file is.
• 06:33Where is he
• 06:36wrong? Folder
• 06:40is this guy. Multi image, right?
• 06:42And so there's three images
• 06:44in that CDI file, right?
• 06:46And you're seeing three
• 06:47different images in the file.
• 06:48Now you don't have to
• 06:49convert every single image.
• 06:50So the nice thing about this,
• 06:51importing the data into an arena
• 06:53that has a data set like this.
• 06:54I can right click just on image
• 06:56#2 and just convert image #2.
• 06:58I don't have to convert image #3 and
• 07:00image #1 because maybe I don't want them.
• 07:02Maybe there were bad images,
• 07:03maybe there was something wrong with them.
• 07:05Maybe you don't really want
• 07:06to convert those files.
• 07:07You can just convert this file.
• 07:08If you want to convert multiple files here,
• 07:10you can select all three of them,
• 07:12hit right click and convert those
• 07:14files into the native format.
• 07:16So that's the main idea and
• 07:17the idea of doing it this way
• 07:19with the raw microscope format.
• 07:21It's the ideal way of doing it is that not
• 07:23only are you converting it inside the Mrs.
• 07:25software itself,
• 07:26but it's also converting for the most part,
• 07:29usually converts the metadata.
• 07:31The metadata contains the channel colors,
• 07:33it contains the voxel sizes.
• 07:35And it makes sure that when you
• 07:38open this file inside of Mrs.
• 07:39and you come in here and you want to
• 07:41make a measurement from here to here.
• 07:43Say I come in here and I want to make
• 07:45a measurement from here to here,
• 07:46there's a measurement tool over.
• 07:47It's just a quick little measurement tool.
• 07:48That measurement tool is going to make sense.
• 07:50So if I come up here and I try to
• 07:52measure one of these little nuclei here,
• 07:54it's telling me that that's
• 07:55five microns across.
• 07:56So that means the voxel sizes are
• 07:58are properly imported in in the
• 08:00measurements and all the measurements
• 08:01and statistics that you end up getting,
• 08:03whether it's volume,
• 08:05surface area,
• 08:06any other parameters that involve the
• 08:08the length or the shape or the size are
• 08:11going to be properly converted into
• 08:12the right format for Mars processing.
• 08:15So anyway,
• 08:15I'm going to stop there.
• 08:16That's just kind of the general
• 08:17idea of kind of making sure that
• 08:19your files are converted right.
• 08:20Now there is a second way of converting data.
• 08:23There is a stand alone file converter
• 08:25and I'll launch it over here.
• 08:27It's going to look something like this.
• 08:28Now,
• 08:28I'm not going to go into a
• 08:29whole lot of details here.
• 08:30If you have some problems importing
• 08:32your files or you're not sure
• 08:33if it's not working right,
• 08:34or if you've converted it and it
• 08:36doesn't convert right, let us know.
• 08:38Sometimes there's bugs,
• 08:39sometimes there's file formats
• 08:40that we don't support.
• 08:41Sometimes there's issues with
• 08:42different types of file formats
• 08:43or the way you acquire your data.
• 08:45Just let us let us know.
• 08:46Contact the support team,
• 08:47we can help you do that.
• 08:49But the file converter is nice because
• 08:51you can put this file converter anywhere.
• 08:52It doesn't have to be on the Mrs.
• 08:54workstation. You can put it on your laptop,
• 08:55You can put it on your office machine,
• 08:57you can put it on your
• 08:58acquisition machine if you wish.
• 09:00You put it anywhere you want.
• 09:01So as soon as you require your data,
• 09:02you can actually start the
• 09:04conversion process Here, however,
• 09:05you're just going to drag your
• 09:06files in or add your files and
• 09:08you'll have the parameters here.
• 09:09And then you'll be there's a start
• 09:11button and you can convert those files.
• 09:13Now this is a little bit different
• 09:14than the Arena view in the sense that
• 09:15this is going to convert everything.
• 09:17You don't have an option of just
• 09:18converting one out of those three
• 09:19size files that I showed you earlier.
• 09:21It's going to either do all or not.
• 09:23But to make sure that this
• 09:25works at all times,
• 09:26you have to set a voxel size.
• 09:28So if for whatever reason the files that
• 09:30you require don't have a voxel size,
• 09:32it will ask you for that voxel size.
• 09:34Now if you don't know the boxel size
• 09:36at the time, you can still convert it.
• 09:37You can put a dummy value in
• 09:39there and change it later,
• 09:40but the idea is that you have to.
• 09:42The idea is that you know the voxel size,
• 09:44the voxel parameters XY and Z,
• 09:46so they input them in there and
• 09:47then when you convert the file
• 09:49it's ready to be used immediately.
• 09:51And As for TIF files,
• 09:52I'm not going to get into the
• 09:53TIF files very much here,
• 09:54but at the TIF files we have
• 09:56a TIF reading series.
• 09:57So whenever you have a TIF stack,
• 09:59they're named in a particular convention.
• 10:01There is an option here under
• 10:03settings to really modify those
• 10:05TIF parameters to make sure you
• 10:06have the proper reading frame
• 10:07that how many channels you have,
• 10:09how many, how many sections you have,
• 10:12and things like that.
• 10:13You can kind of manipulate it a
• 10:15little bit to make sure that it's
• 10:16kind of compiled in the right way.
• 10:18Now, if you name your files properly,
• 10:19this works like a dream and it's really nice.
• 10:21So if you have any questions on
• 10:23saving files that's TIFF and make sure
• 10:24you input them into Mars properly,
• 10:25let us know.
• 10:26I have a separate video tutorial
• 10:28that I can I can share with you
• 10:30later that talks about how to import
• 10:32TIFF files into the software.
• 10:33OK, so let's close that up. OK.
• 10:35So let's jump right into Amaris here.
• 10:37So this is not the file I'm
• 10:38going to use today.
• 10:39So today I'm going to focus on
• 10:41maybe two or three different
• 10:42files that I'm going to show you
• 10:43that can kind of cover a lot of
• 10:45the different features in Amaris.
• 10:48This file here is one of my favorites.
• 10:49This is a Microglia data set that has
• 10:53a really nice red stain and it also
• 10:56has a secondary marker that's called CD 86,
• 10:58which is an activation protein for
• 11:00the Microglia. And so there's a lot
• 11:02of cool localization between this
• 11:04red channel and this green channel.
• 11:05So the red channel looks like this, right?
• 11:08You can see a little bit of
• 11:09background in the sample here.
• 11:11It's not perfectly clean,
• 11:12which is really nice.
• 11:12Kind of gives us a lot of places
• 11:15to work with in terms of making
• 11:17sure that we can render objects
• 11:19that are not perfectly clean.
• 11:20Where there's a little bit of auto
• 11:22fluorescence or background or tissue
• 11:24auto fluorescence or any kind of
• 11:25tissue background or junk or dirt
• 11:27in the data set in the channel.
• 11:28And then there's this green
• 11:29channel that has a bunch of little
• 11:31punked up and we're going to,
• 11:32we're going to kind of do some rendering
• 11:34and quantification of this function
• 11:36in that various different ways.
• 11:38Now the first thing we're going to
• 11:39talk about a little bit about is
• 11:40the navigation inside the software.
• 11:41Again,
• 11:42I'm going to be doing this as we go here.
• 11:44And so the idea is that,
• 11:45you know a lot of this is
• 11:46kind of going to be the same.
• 11:47So let's look at this image here.
• 11:49So the basic manipulations
• 11:50inside the software,
• 11:52when you're looking at just the volume view,
• 11:54you have this,
• 11:54if you left click and rotate,
• 11:56you can turn your image around in 3D.
• 11:58This reset button straightens it
• 12:00back to the primary default position
• 12:02in the way that it was acquired.
• 12:05If you hit the mouse wheel
• 12:06in the middle here,
• 12:07you can zoom in anywhere where your mouse is.
• 12:09If I'm in the center there, it'll zoom there.
• 12:10If I have my mouse over here,
• 12:12it'll zoom into my mouse over here as well.
• 12:14And then you can right click and you
• 12:16can pan your image around to kind
• 12:18of identify where that sample is.
• 12:19This is our 3D view.
• 12:21This is the kind of the primary
• 12:23view where you're going to do the
• 12:24majority of your work instead of Mars.
• 12:26Sometimes people like to look at
• 12:27it in a slice view just to kind
• 12:29of navigate and look at their
• 12:30data a little bit more closely.
• 12:31I use this very often to try
• 12:33to make sure and see,
• 12:35make sure we have the proper resolution
• 12:37to quantify the objects of interest.
• 12:39Sometimes the biggest problem I have
• 12:40in the support team is that people
• 12:42send me data and they're trying
• 12:43to render something that's really,
• 12:45really small, but they've taken a really,
• 12:47really large step size.
• 12:48And that makes it really difficult
• 12:49to render something in 3D if you
• 12:51don't have the proper resolution
• 12:52to kind of render that object.
• 12:53And we'll talk a little bit about
• 12:54that as we move a little bit forward.
• 12:56But that's really an important
• 12:57thing to keep in mind before you
• 12:59bring your data into a Morris
• 13:00know what you're trying to render.
• 13:02If you're trying to render something
• 13:03that's about 1/2 a Micron in diameter
• 13:05like these little fibers here,
• 13:07these little processes of this microglia,
• 13:08that's about 1/2 a Micron,
• 13:101 Micron in diameter and you're taking A2
• 13:12Micron step size or something really large,
• 13:15you're not going to get a really good
• 13:163D rendering of that object because
• 13:17your step size is it's not really
• 13:21conducive to kind of rendering this
• 13:23guy in 3D because you're taking it.
• 13:24You're taking too big of a step size.
• 13:26So always keep that in mind when
• 13:27you're before you kind of go into
• 13:29the microscope and acquire images,
• 13:31Do you have the right step size,
• 13:32how good can you go?
• 13:33Can you make it the proper step size.
• 13:35Most microscope have a button on the system
• 13:38where you can say optimize your Z step,
• 13:40C step is really important
• 13:42especially in 3D rendering.
• 13:43And that optimize these steps gives you
• 13:45what's called Nyquist level sampling
• 13:46and that's going to give you the best
• 13:48that your acquisition system can achieve
• 13:50with the objective and the optics that
• 13:52you have currently set up on that system.
• 13:54And so that's something that I
• 13:56always try to recommend people to do.
• 13:57It's not always feasible.
• 13:58Sometimes it's going to
• 13:59give you too many steps,
• 14:00sometimes it's going to take too long.
• 14:01You can,
• 14:02you don't have to use that button and set
• 14:04that optimal step size every single time.
• 14:06But if you're trying to get the
• 14:08best possible image with the
• 14:09best possible Co localization,
• 14:10rendering and that sort of stuff,
• 14:12you definitely want to try to kind
• 14:14of keep using that that optimal
• 14:15step size as best you possibly can.
• 14:17And again,
• 14:17there's times where you don't have to.
• 14:19But again,
• 14:19if you're rendering something that's
• 14:21small and a lot of times people are,
• 14:23you want to try to keep it as
• 14:25optimal as you can and to to render
• 14:27those objects as best you can.
• 14:29So again, that's our slice review.
• 14:32You can look at 1 channel, 2 channels.
• 14:33This is our display adjustment.
• 14:34You can adjust the levels here.
• 14:36I did that already. This is just a display.
• 14:38You're not removing the channels.
• 14:40It's just a display adjustment to kind
• 14:41of remove some of those lower pixels.
• 14:43So your signal looks a little bit blacker.
• 14:45So you can see your structure
• 14:46a little bit better.
• 14:46And you can do this at any point.
• 14:48And you can take a snapshot of these
• 14:50images at any point you adjust your display.
• 14:53There's a snapshot button up here.
• 14:55So if you like this image,
• 14:55you want to take a snapshot of this image.
• 14:57You can move this scale bar over here,
• 14:59put it in your image and you
• 15:00can hit take snapshot.
• 15:01Now now usually I set this
• 15:03snapshot to save to my desktop.
• 15:04You can choose wherever you wish to save it.
• 15:07Usually I save it as a TIFF image.
• 15:09We do have a couple other options here,
• 15:10but I never save my tips
• 15:12my screenshots as those.
• 15:14They're usually a little bit smaller
• 15:15and they're somewhere compressed.
• 15:16TIFF images are a little bit better.
• 15:17It maintains that the true resolution
• 15:19of the of the image on the screen.
• 15:21And so if you click this and take a snapshot,
• 15:23it'll automatically take a snapshot
• 15:24and you have a nice high resolution
• 15:27snapshot of that structure
• 15:28of exactly what we're looking at.
• 15:29Again, I have those little lines in there.
• 15:31So whatever you see on the screen
• 15:32that is going to be in the snapshot.
• 15:34So that snapshot just so you know has
• 15:36those little 2 little squares in it.
• 15:38So if you don't want those
• 15:39little squares over here,
• 15:40obviously delete those structures
• 15:41over here and then take that snapshot.
• 15:43So whatever you see on the screen is
• 15:45what you get in terms of your snapshot.
• 15:47So always remember that.
• 15:48Same thing goes when we,
• 15:49when we talk a little bit
• 15:50about the animation later.
• 15:51Whatever you see on the screen,
• 15:52when you're setting your animation up,
• 15:53that's what's going to be recorded
• 15:54and we'll do a quick animation towards
• 15:56the end of the end of this session.
• 15:58Just to give you an idea of how that works.
• 15:59Because that's a really cool little
• 16:01feature for lab presentations,
• 16:02making quick little movies,
• 16:04being able to kind of identify and
• 16:06show something to a colleague to say,
• 16:08hey, this looks really cool and make
• 16:10a nice little 3D movie of that makes
• 16:12it really simple and easy to do.
• 16:14So another view.
• 16:15Here is our section view.
• 16:16Again,
• 16:16this is mostly just a visualization tool.
• 16:19If we look at this guy here you have
• 16:21your XY view, You have your two X,
• 16:23you have your XZ and your XY views.
• 16:25You can move this look I can
• 16:27click on where where is that?
• 16:36There it is. Oops, there it is.
• 16:38It's bigger and you can zoom in here
• 16:40and you can adjust the slicer here.
• 16:42You can click your thing here and
• 16:43you can kind of move exactly in here.
• 16:45So if you were kind of interested
• 16:46in what was going on right there,
• 16:48I can click my cursor right
• 16:49there and you can see it's right
• 16:51smack dab on the structure there.
• 16:52If you want to get rid of those crosshairs,
• 16:54you can get rid of those crosshairs as well.
• 16:58Come up here and you can turn those
• 17:00crosshairs off and you can see exactly
• 17:01exactly where you're going in those sections.
• 17:03So just another visualization again,
• 17:05you can take a snapshot of
• 17:06any of these views as well.
• 17:08And there's an extended section here.
• 17:10So you can that that crosshair here.
• 17:13You can make it a little bit wider and
• 17:15show more of your data in this way,
• 17:18in this way as well.
• 17:19And you can show a little bit more of
• 17:21the data if you want to take a snapshot
• 17:23that's in showing the data inside
• 17:24of that little region of interest.
• 17:25There's a lot of different options here.
• 17:27For visualization,
• 17:27let's go back to our 3D view here.
• 17:30So let me adjust my levels here a little bit.
• 17:36So we're going to focus today on the
• 17:38green and the red channel and we're
• 17:40going to focus on making surface
• 17:42renderings first of the structures.
• 17:43Now there's a lot of different
• 17:45ways to make surfaces in a Mars.
• 17:47I've I've done a bunch of them
• 17:49already here and I'm going to talk
• 17:50a little bit about all three of
• 17:52them and I'm going to just talk very
• 17:54briefly about what are why one,
• 17:55why we, why would you choose one
• 17:57over the other, excuse me,
• 17:59why you would choose one over the other.
• 18:02And the way we've done surfaces for
• 18:04a long time in a Mars for the last 30
• 18:07odd years we've done what's called
• 18:10an intensity based surface creation.
• 18:12Basically what that means is we do a
• 18:14little bit of pre processing of our data.
• 18:16We identify that the structures
• 18:19are a particular size.
• 18:21There's some background.
• 18:21We do a little bit of smoothing,
• 18:23do a Gaussian blur,
• 18:24which is very traditional way
• 18:26people do segmentation,
• 18:28whether it's in Fiji or
• 18:29other software applications,
• 18:30they're very typical first step to
• 18:32kind of clean up your data is to do
• 18:35these kind of smoothing activities.
• 18:36Now recently, over the last 5-6 years,
• 18:39microscope companies have done
• 18:40their own smoothing and they do
• 18:42their own background subtraction.
• 18:43They, you know, Leica has Thunder,
• 18:45Zeiss has Airy scan, Nikon has denoising.
• 18:49There's all these really fancy tools
• 18:51that are doing these preprocessing
• 18:54already in the acquisition system.
• 18:55It's all post processing,
• 18:56but it's all after they acquire
• 18:58the data and it's all based on
• 18:59their optics of the microscope.
• 19:00And they're trying to optimize the
• 19:02visualization to show the signal,
• 19:04what is signal and what is noise
• 19:05and kind of get rid of as much
• 19:06of that noise as we can using a
• 19:08lot of those features.
• 19:09They actually do a lot of smoothing already.
• 19:12And so sometimes you don't have to do
• 19:13any blurring in the Morris to do that.
• 19:15And again it's something that you
• 19:16need to be made aware of because a
• 19:18lot of times the biggest pitfall
• 19:20people run into with the Morris
• 19:21is they come in here,
• 19:22I'm going to just rebuild this surface here,
• 19:25right.
• 19:25The the most common problem that
• 19:27people run into is that they run
• 19:29through a Mars and we're wizard based
• 19:32as we're going to see a lot of the
• 19:34tools we're going to come in here to
• 19:35make these surfaces and make these
• 19:37renderings are all wizard based.
• 19:38We're going to kind of step through
• 19:40the process within the wizard.
• 19:42Inside the wizard,
• 19:43you're going to make some choices,
• 19:44whether it's smoothing,
• 19:45whether it's background subtraction,
• 19:46whether it's a threshold splitting,
• 19:48a couple of different features.
• 19:49There's not a lot of options here.
• 19:51We try to limit it to make it
• 19:53as simple as possible,
• 19:54but the biggest pitfall people
• 19:55run into is they just use the
• 19:57default parameters and that's it.
• 19:59They come in through here,
• 20:01they'll create this little blue BLOB up here.
• 20:03Right. And you'll see surfaces,
• 20:04one get created and they're
• 20:05going to walk through the wizard.
• 20:06You'll see a couple things that are checked.
• 20:08Sometimes you don't need these
• 20:09features and you turn them off, right.
• 20:10I'll talk a little bit about what
• 20:12these features are a little bit later,
• 20:13but usually I turn them off until
• 20:14I know that I'm going to use them.
• 20:16Always turn them on later within the process.
• 20:19But the idea is that the biggest part
• 20:21of a Mars is you have to be able
• 20:23to render the objects of interest.
• 20:25If you can't do that,
• 20:26whether it's with a spot object,
• 20:27surface object or filament object,
• 20:29you have to kind of find maybe
• 20:31another way to label the structures.
• 20:33Maybe they're not labeled well enough.
• 20:34Maybe you have to improve the antibody
• 20:36stain and get rid of the background.
• 20:37There's a lot of different things you can
• 20:39do to kind of make the software work better.
• 20:41But the idea is that we need to be able
• 20:42to segment the objects of interest.
• 20:44Now when you get into this particular
• 20:46step here, we're at step number two,
• 20:47first step, there's some,
• 20:48there is a region of interest tool here.
• 20:51Not going to get into a whole
• 20:52lot of ideas here,
• 20:52but this region of interest,
• 20:54basically if you did this,
• 20:55you would be able to kind of choose
• 20:57a little box region here.
• 20:59You can adjust the size here and
• 21:00you can analyze.
• 21:01It's just a little box the the limitation
• 21:04here is that it has to be a cube.
• 21:07There are there are options to do
• 21:08kind of non cubed region of interests
• 21:10and again if that's something that
• 21:12your data requires contact support,
• 21:14we can walk you through some some ways
• 21:16to kind of analyze bits and pieces.
• 21:17That's not a square,
• 21:19but for today we're going to process
• 21:21the entire image here because we have
• 21:23all these little microglia that we
• 21:25want to render as a structural object.
• 21:27So a lot of the things that we're going
• 21:29to do here require that smoothing factor,
• 21:32background subtraction or the
• 21:33new feature in a Mars 10,
• 21:35which I'm going to 10.1 rather is
• 21:37this machine learning segmentation,
• 21:38which is I'm going to walk through
• 21:39that with the most detail here.
• 21:41But what I wanted to just kind of
• 21:43set the stage for is that the the
• 21:44most common thing that people end up
• 21:46doing is they keep this the default,
• 21:48which this value here is double
• 21:50the voxel size,
• 21:51that that is just a default value.
• 21:52Don't get stuck by using that value
• 21:54because that is not always the right value.
• 21:56Sometimes it's too much smoothing
• 21:58and it blurs the data and you lose
• 22:01detail on your spine,
• 22:02small little structures.
• 22:03Now if you're doing cell bodies
• 22:04and big structures,
• 22:05the smoothing factor is going
• 22:06to be perfectly fine,
• 22:07but you start looking at small
• 22:09fine structures of dendrites,
• 22:10processes, little fibers,
• 22:12fibers like structures.
• 22:13You're not going to want
• 22:15to have too much smoothing.
• 22:16You should have some smoothing whether
• 22:18you do it on your acquisition side
• 22:21using denoising or what have you.
• 22:23You want to kind of improve
• 22:24that a little bit here.
• 22:24And usually I go to about
• 22:26a single pixel width,
• 22:27especially when I'm doing
• 22:28something that's really fine
• 22:29structure like these dendrites,
• 22:31like there's a lot of fine structures here.
• 22:32These guys are not very big,
• 22:34they're only maybe at some points
• 22:35only like two or three pixels across.
• 22:37So we want to keep that relatively
• 22:39low and this is a common practice
• 22:40no matter which method we're using,
• 22:42whether you're going to use machine
• 22:43learning or you're going to use the kind
• 22:45of the general intensity based creation.
• 22:47Both of them are really important to make
• 22:49sure that we're getting the detail that
• 22:50we want to render within our objects.
• 22:52Because if I that is too high,
• 22:54what's going to happen is we're
• 22:56going to start blurring this data
• 22:58and our surface is going to include
• 23:00spaces in between fibers like this
• 23:01where this data in the middle of
• 23:03this little structure here is going
• 23:04to be created as part of the surface
• 23:06structure where in reality it's not.
• 23:07It's a branch that comes up here
• 23:08and goes this way as we want to make
• 23:10sure that we can render as much as
• 23:11we can and visualize what we can.
• 23:13Again, always goes back to the acquisition.
• 23:14The better your acquisition,
• 23:15the better we're going to be able
• 23:17to kind of determine where that
• 23:18fiber is and where that fiber isn't,
• 23:20whether it's in 3D and in two D as
• 23:22well along the XY axis to make sure
• 23:24that we have the resolution there.
• 23:26And X&Y because sometimes people
• 23:28take an image at 512 by 512 and
• 23:30they're trying to render small
• 23:31little puncta and it's only defined
• 23:33by one or two pixels across.
• 23:35We can render it,
• 23:36we can identify it,
• 23:36we can render it,
• 23:37but we can render it better if it
• 23:39was 6 or 8 pixels across.
• 23:40And the higher resolution,
• 23:41we can really get a really fine
• 23:43control and measurement of the volume
• 23:45of those structures if we have more
• 23:47pixels to kind of render that object.
• 23:48And so that's all we're doing
• 23:50here with these smoothing factors
• 23:51and this background subtraction.
• 23:53These are tools we can use to kind of
• 23:55identify the labeling of the structure.
• 23:57And like I said,
• 23:58I'm not going to walk through a
• 23:59lot of the details specifically,
• 24:01but I do have these guys already
• 24:02created here.
• 24:03So this here is the basic intensity
• 24:06based surface structure.
• 24:07I'm going to zoom out a little bit here.
• 24:11Oh, you know what?
• 24:11I think I redid it here.
• 24:12Let me just let me just go through the
• 24:13wizard here just to show the final result.
• 24:15You can see how fast it is here.
• 24:16It's going to take this image here.
• 24:18It's not a super small data set,
• 24:19but it's also not super big.
• 24:20I'm going to change the color here so
• 24:22we can see it a little bit better.
• 24:23And so the visualization here
• 24:25in the surface was two ways.
• 24:26We talked about the 3D view and the
• 24:28slice view, kind of in a general sense,
• 24:30but in the 3D view,
• 24:31we can switch it to this.
• 24:33This is our 3D view and you
• 24:35can see the structure of these
• 24:37Microglia that we have created here.
• 24:39But we also have this little box
• 24:40here that switches to our slicer.
• 24:42It's a really powerful little tool in Mars.
• 24:45As you zoom into your data set
• 24:46during the creation process,
• 24:48where you're creating the border
• 24:49and selecting the pixels that
• 24:50make up your surface,
• 24:51you can visualize this outline.
• 24:53And this outline has a little
• 24:55there's a little ball chain here,
• 24:58Let's see if I can move it.
• 25:00If I click on this and
• 25:01move it left and right,
• 25:02I can go up and down through
• 25:03the Z stack and I can say,
• 25:04hey,
• 25:04those are the surfaces that were created.
• 25:06You can see the surface that
• 25:08we're rendering as based on
• 25:09the border of the structure.
• 25:11And so as we kind of go up and
• 25:12down and look at the surface here,
• 25:13you can see how it's rendering this
• 25:15structure and you can see exactly what's
• 25:17being included and what's not being included.
• 25:19Like for example,
• 25:20you can look at this structure
• 25:21right here and you can see it's,
• 25:23it's not and it's including this little
• 25:26area in between there as one big structure,
• 25:30right.
• 25:30It's not seeing this as a
• 25:31fiber here and a fiber here.
• 25:32Now that's partly based on the
• 25:33fact that the way I'm making
• 25:35this particular surface is using
• 25:37a broad intensity based surface
• 25:38with no background subtraction,
• 25:40just using a smoothing factor kind
• 25:41of create this surface and that's OK.
• 25:43There's nothing wrong with that.
• 25:44Obviously,
• 25:44you know you're going to get this
• 25:46little bit of an error here,
• 25:47but the rest of it works pretty
• 25:48good and you can get a good
• 25:49rest estimation of your volume.
• 25:50But you can see there's areas here
• 25:52where there's signal where maybe
• 25:54there shouldn't be any signal, right?
• 25:56We should not be selecting that.
• 25:57But because we did the smoothing,
• 25:58we set the threshold relatively low,
• 26:00We want to kind of make this surface,
• 26:02we have these kind of inconsistencies
• 26:03in some of these surfaces.
• 26:05And it just depends on the data set.
• 26:07Sometimes it's a really nice way to do it,
• 26:08especially if you have a nice
• 26:10thick Soma in this particular
• 26:11example that's labeled really well.
• 26:13You know, it's not, there's no gaps,
• 26:15there's no holes,
• 26:16there's no anything in there,
• 26:17and it works really well.
• 26:19The second method is to do the same sort
• 26:21of thing with a little bit of smoothing,
• 26:24but we're going to use what's
• 26:25called background subtraction,
• 26:26local background subtraction.
• 26:27Again, I'm just going to just,
• 26:29I'm going to zoom in here,
• 26:30let me make this white so we can
• 26:32see a little bit better, right?
• 26:34I'm going to zoom in here and that same
• 26:37area here where we were looking at before.
• 26:39Now you can see I'm able to kind of identify
• 26:42these little structures all by themselves.
• 26:44Now I'm not including these areas.
• 26:47So now these little areas
• 26:48here are not being included.
• 26:49Whereas if I go into this guy here,
• 26:53we can look at these guys simultaneously,
• 26:56see how different they are.
• 26:57Let me change the color a little bit here.
• 26:59This is a good example.
• 27:00Make it pink, purple.
• 27:02Here.
• 27:03You can try to see exactly
• 27:04where your border is.
• 27:05You can see that they're not quite the same,
• 27:07right?
• 27:07Because now with the background subtraction,
• 27:10we're able to kind of get a little bit
• 27:11closer to the edge of the structure,
• 27:13whereas with the smoothing only you can
• 27:15see it's usually a little bit bigger,
• 27:17little bit broader.
• 27:18And the structure here,
• 27:19it's selecting more of those
• 27:21pictures on the edge within the
• 27:23structure and we might not be able
• 27:25to kind of get those small little
• 27:27structures like even in here you
• 27:28can see we're able to kind of get
• 27:30these structures a little bit finer,
• 27:31a little bit better control
• 27:33where those edges are.
• 27:35And so that's the second method
• 27:36of making these surfaces using
• 27:37that background subtraction.
• 27:39Again, background subtraction.
• 27:40And if we look at it in the
• 27:43wizard here is this value here.
• 27:45And it's just basically the
• 27:46way I like to think about it,
• 27:47is it's the size of the object
• 27:49you're trying to render.
• 27:50So if your object is about
• 27:51half a Micron in diameter,
• 27:53you're going to set this to half a Micron.
• 27:54If it's about 1 Micron,
• 27:55you're going to set it to 1 Micron.
• 27:57And it does a local background subtraction.
• 27:59So that's why it helps
• 28:00identify these structures.
• 28:01Now what I like to do,
• 28:03and this is a good rule of thumb for a
• 28:04lot of new users because you're like,
• 28:06what did these actually do to my data?
• 28:08Because you can see the image
• 28:09here hasn't changed when I'm
• 28:11processing these these values here,
• 28:13you don't see the image change on
• 28:14the screen and it's not going to when
• 28:16you're processing it through the wizard.
• 28:17But you might ask yourself what does a
• 28:20smoothing factor of .13 do to my data?
• 28:23Because that's the image,
• 28:24that's the threshold.
• 28:25When we get to this next step
• 28:26here in the wizard,
• 28:27this threshold that that's generated here
• 28:29is based on that background subtraction,
• 28:31that smoothing whatever values
• 28:32you put into that, into those,
• 28:34into those algorithms.
• 28:36And so the idea is that what is
• 28:38that actually doing to my data set?
• 28:40So what I like to show people a lot of
• 28:41times is I go to this image proc window
• 28:43just for just for a preview, right.
• 28:45And the idea is that the top
• 28:48view is your original data,
• 28:49the bottom view is the processed data.
• 28:52So I'm going to switch this to
• 28:54slice view kind of like to see
• 28:55it in the best possible way.
• 28:56I'm going to zoom in here actually.
• 28:58Let me do that one more time.
• 28:59I did that wrong. Sorry.
• 29:02Close that up. Image proc slice view.
• 29:05Here we go. Right.
• 29:07So you have these two images
• 29:09here that they're identical.
• 29:10The the the image on the top and the
• 29:12image on the bottom are they should be
• 29:14identical and they're not identical.
• 29:22Here we go. So the top and the
• 29:24bottom are identical, right?
• 29:26Visually, if I come up here and I
• 29:29click on the Gaussian filter and I
• 29:31do .13 which is a single pixel width,
• 29:34you can see the image at the top,
• 29:35which is your original image,
• 29:37it starts to look like the
• 29:38image at the bottom.
• 29:38You can see it gets a little bit blurry,
• 29:40all these pixels.
• 29:41That basically what a Gaussian blur does
• 29:43is it reduces the pixel variation so you
• 29:45get a little bit of a smoother transition.
• 29:47So again, all those post processing tools
• 29:49out there that acquisition systems do,
• 29:51that's essentially kind
• 29:51of what they're doing,
• 29:52a little bit fancier algorithm.
• 29:53But Gaussian smoothing blur is kind of
• 29:55the default for years to kind of blur
• 29:58that image so that we have a little bit,
• 30:00way better way of selecting the
• 30:02pixels that are part of our structure.
• 30:04Now that is the gusting point.
• 30:06Now, if I were to take this and say,
• 30:07hey, I want to, if I set this to like .3,
• 30:10which is closer to double the box slide,
• 30:11all of a sudden now things get
• 30:12a little bit blurry here.
• 30:14Now we don't see this inside
• 30:15a little bit much.
• 30:16We don't see the edges clear.
• 30:17It's kind of blurring that edge a little,
• 30:19a little bit.
• 30:20Especially when you have another
• 30:21fluorescence in your tissue,
• 30:23you're going to get a blurrier edge
• 30:25around your structure a little bit.
• 30:26And so again,
• 30:27as you go higher and higher here,
• 30:28obviously you're going to get
• 30:29blurrier and blurrier, right?
• 30:30The more you go,
• 30:31the blurrier it's going to get.
• 30:32Obviously,
• 30:32you're not going to go this side
• 30:33because now all of a sudden you
• 30:34can kind of see your the structure
• 30:35is still there, but again very,
• 30:37very little detail is there.
• 30:38So you're not going to make a
• 30:39surface to something like that.
• 30:40But again,
• 30:40there are cases where you might
• 30:42go high on the smoothing just to
• 30:43kind of make a surface rendering
• 30:44of something that maybe is not
• 30:46labeled perfectly uniformly.
• 30:47Sometimes blurring helps kind of
• 30:48render that objects as a usually
• 30:50for larger object works OK.
• 30:51So I'm going to come back down here.
• 30:53We're going to set this back to .25,
• 30:55something like that.
• 30:55And so the idea is that if you
• 30:57do a little bit of smoothing,
• 30:58the next step that we have is
• 31:00that background subtraction.
• 31:01So that background subtraction
• 31:02is this guy here.
• 31:04So this background subtraction,
• 31:06again, usually the defaults,
• 31:08it comes out to something
• 31:10usually pretty small,
• 31:11but the software doesn't really
• 31:11know what you're trying to render.
• 31:13Could be something big.
• 31:13It could be something small.
• 31:14Doesn't really know.
• 31:15So don't use these numbers just by default.
• 31:18These guys are about a Micron,
• 31:19Micron and 1/2 in diameter.
• 31:21So I'm going to set this to like
• 31:231.5 and look what happens when
• 31:24you set it to 1.5.
• 31:25All of a sudden now your image
• 31:27looks like this.
• 31:27Now again, the adjustment is a little
• 31:29bit off here, so we can adjust it.
• 31:31We can look at the adjustment,
• 31:32but now all of a sudden now we've
• 31:34cleaned up all this auto fluorescence,
• 31:36all this little, all these pixels in here,
• 31:38they've been kind of reduced to next
• 31:40to nothing because it's background.
• 31:41It does this local background subtraction
• 31:43pixel by pixel across the data set.
• 31:45What that does, again depending
• 31:46on what you're trying to render,
• 31:48what that does is it allows you to
• 31:50find the edge of this surface and
• 31:52the edge of this surface relative
• 31:53to the surrounding background.
• 31:55And so when you go to make a
• 31:57surface and select the pixels that
• 31:58are part of this surface,
• 32:00this is what it's going to look for.
• 32:02These pixels are going to be really,
• 32:03really dim.
• 32:03These pixels are going to be really,
• 32:04really bright.
• 32:05It's going to be really easy to
• 32:06identify those pixels from these pixels.
• 32:08Same thing with this little
• 32:09thing in the middle here,
• 32:10all of a sudden now we have some pixels
• 32:11in there that are clearly really dark.
• 32:13Here.
• 32:13It's mostly background and there
• 32:15are some bright pixels in there.
• 32:16If I didn't do a whole lot
• 32:18of smoothing there,
• 32:18I'm going to select some of those pixels.
• 32:20It's going to be a little bit
• 32:21broader than what it needs to be.
• 32:22So again, the pixel classification here,
• 32:25the background subtraction depends
• 32:26and you can see as I make that small,
• 32:28you see things get a little bit smaller
• 32:30because it's looking at a smaller,
• 32:32it's doing a smaller background subtraction
• 32:33looking at that what that background is.
• 32:35So it's identifying those structures.
• 32:37And again,
• 32:37it's a single threshold value here.
• 32:39But again,
• 32:39if you know that the fibers are
• 32:41all about the same diameter or
• 32:43approximately the same diameter,
• 32:44so this is a nice way to kind of remove
• 32:46this pixelated noise with all these
• 32:47bright pixels out here that we really,
• 32:49don't really care for.
• 32:50We don't want to render them
• 32:51when we get to the next step.
• 32:53And so those are, that's what it's doing.
• 32:54So I do that and I scan a lot of it.
• 32:55It's like, Oh yeah,
• 32:56I'm going to set this to 1:00,
• 32:57I'm going to set this to whatever,
• 32:58and it gives me an idea of exactly
• 33:00what it's doing to my data set.
• 33:01And so those are the two methods for kind
• 33:03of doing kind of a testy base creation.
• 33:05Now I I want to spend a little bit of
• 33:07time on the new feature of a Morris
• 33:09that is using pixel classification.
• 33:10So the idea of pixel classification,
• 33:13it combines a little bit of both kind
• 33:16of the blurring of the image and the
• 33:20creation of kind of a fine structure
• 33:22around the edge of where your signal
• 33:24is and where your background is.
• 33:26But the difference here is
• 33:27that you're going to train,
• 33:29you're going to identify
• 33:30based on selecting pixels.
• 33:32So not based on a blur,
• 33:33not based on anything else.
• 33:34You're going to go in there and say, hey,
• 33:36you know what these pixels are Background,
• 33:38these pixels are signal,
• 33:39make my surface based on that.
• 33:41And so I'm going to walk you
• 33:42through that real quick as to what
• 33:44I did here within this image.
• 33:45I already kind of did it with this image.
• 33:46So I'm not going to read,
• 33:47I'm not going to reinvent the wheel
• 33:49here to kind of show you how it works,
• 33:50but I'm going to show you,
• 33:51you're going to be able to see exactly
• 33:52where I drew within the data set.
• 33:53And again, we're going to kind
• 33:54of just walk through here.
• 33:55We're going to choose our
• 33:56green or red channel.
• 33:57I didn't do any smoothing here.
• 33:59You could do some smoothing.
• 34:01The smoothing is a post smoothing
• 34:03process of the end result,
• 34:05but again,
• 34:06when I do fine structures I
• 34:07often do turn it off again.
• 34:09It just depends on the data set
• 34:10whether you're going to have this
• 34:11turned on or turned on again.
• 34:12When I'm doing nice fine structures,
• 34:14punctive processes often,
• 34:16especially for the the
• 34:18machine learning part of it,
• 34:20I do sometimes turn this off.
• 34:21Now for the the intensity base,
• 34:23I rarely turn the smoothing off.
• 34:24It's almost always on in some way,
• 34:27shape or form.
• 34:27But we're going to come in here.
• 34:29We're going to do the machine learning.
• 34:30I have two options with machine learning.
• 34:31One is you can do it on the red channel by
• 34:34itself or you can do it on all channels.
• 34:35Now why would you ever want
• 34:36to do it on all channels?
• 34:38All channels would be a a data set.
• 34:40For example,
• 34:40if you had a bright field data set
• 34:42that was RGB and you had maybe
• 34:45H&E stain or something like that,
• 34:47you would kind of do the machine
• 34:48learning on all three channels.
• 34:50And then you would train your
• 34:51purple cells or your brown cells,
• 34:53whatever is in that sample,
• 34:54to be signal and then everything
• 34:56else would be background.
• 34:57And then you can render cells within it.
• 34:59This kind of light microscopy world,
• 35:01kind of an H&E or type of Histology type
• 35:05slide where you kind of have an RGB channel,
• 35:09but you want to use the machine learning
• 35:11to kind of make the surfaces and
• 35:12count the cells and things like that.
• 35:14That would be a case where
• 35:15you'd use all channels.
• 35:16And there's some cases as well when you're
• 35:18doing fluorescent data imaging where
• 35:19you want it to use all channels as well.
• 35:21In this case,
• 35:22we don't because these are two
• 35:23totally different channels,
• 35:24totally labeled structures.
• 35:25We're just going to do it on the
• 35:27red channel by itself.
• 35:31OK, so the main interface
• 35:33is pretty straightforward.
• 35:34If you've never done any kind
• 35:36of pixel classification before,
• 35:38the concept is pretty simple.
• 35:40You have a foreground and
• 35:42you have a background.
• 35:43You have a draw tool.
• 35:44In this case it's a little circle.
• 35:46You can change the size of the
• 35:48circle using the mouse wheel,
• 35:49holding down the control key
• 35:51and using the the mouse wheel
• 35:52to control that structure.
• 35:54If you do have a secondary
• 35:55mouse wheel on your system,
• 35:57this slicer mode that we have
• 35:59controls this Z slice as well.
• 36:01I didn't mention that earlier,
• 36:02so we do recommend a specific
• 36:03mouse on a requirements page.
• 36:05It's AMX Master 3 from Logitech,
• 36:08but any mouse that essentially
• 36:09has a a secondary mouse wheel
• 36:12can control this slicer.
• 36:13No matter where you are and it's a really,
• 36:14really nice convenient feature,
• 36:16you can also do the keyboard
• 36:17up and down arrows as well.
• 36:19And you also you can always
• 36:20click on this guy and move this
• 36:21guy left and right as well.
• 36:22There's three ways to kind
• 36:23of change those size,
• 36:24but the the mouse wheel is the nicest way.
• 36:26So the idea here is I'm going to
• 36:28try to find where I trained here.
• 36:30So you can see here's some
• 36:33strokes here you're going to see
• 36:34these purple and green strokes.
• 36:36So the purple strokes represent background,
• 36:38the green strokes represent signal.
• 36:40So you draw these strokes and
• 36:42again you just come in here,
• 36:43you're going to change the size of this
• 36:45guy like this and you're going to say,
• 36:46hey,
• 36:47you know what,
• 36:48hold down shift and you're going to
• 36:50draw signal and then you hit train
• 36:51and predict and you're going to train
• 36:53different areas of your data set.
• 36:55Again,
• 36:55that to use pixel classification,
• 36:57the basic rule of thumb is that
• 36:59you're going to train a lot of
• 37:01different areas in your data set
• 37:02that have different types of signal.
• 37:04Dim signal, bright signal,
• 37:05high background, low background,
• 37:07dirt, garbage.
• 37:08Wherever you're going to train,
• 37:10you're going to train this structure and
• 37:12again you're going to train everywhere,
• 37:13the soma, the processes,
• 37:16the background, every everywhere.
• 37:17So if we come up here and say you know
• 37:20what that should be background, right.
• 37:21This is not signal.
• 37:22That's clearly background come up here,
• 37:24click background,
• 37:25come up here and we're going to
• 37:27train it a little bit like that.
• 37:30And you're going to hit train and
• 37:31predict and you're going to see.
• 37:32It's a pretty fast process to take that.
• 37:34Just that one little training we did there.
• 37:36It's going to recalculate everything just
• 37:37based on training that little structure.
• 37:39And all of a sudden now I have
• 37:40a new signal and you can see
• 37:42that may have changed,
• 37:43not just this area here,
• 37:44which now has no signal in that
• 37:46little area that I trained,
• 37:48but it may have changed some other
• 37:49areas as well for the better.
• 37:50Hopefully that's the goal
• 37:51where maybe now it's picking up
• 37:53these little holes in here that
• 37:54it didn't pick up before, right?
• 37:56So you go through the data set,
• 37:57you go through the Z stack
• 37:58and you train the structures.
• 38:00Now if you look at here and you say,
• 38:01OK, well that's training a little
• 38:02bit too wide for my structure,
• 38:04that's OK, come in here and train it,
• 38:06come in here and train along
• 38:07here and draw the structure.
• 38:09Now the way our training works is
• 38:12that we're visualizing this slicer,
• 38:15this data, this volume and what's
• 38:17called an extended section.
• 38:18So down here at the bottom right
• 38:20hand corner you see this in
• 38:22extended section section here,
• 38:24the size of this is the size of that
• 38:27little yellow box you're seeing right here.
• 38:29The bigger you set it,
• 38:30the bigger that structure is going to be,
• 38:32the more you're going to see in terms
• 38:33of the the thickness of your structure.
• 38:36Now, the rule of thumb is that usually
• 38:38you're going to set this to about the
• 38:40size of the object you're trying to render,
• 38:42whether it's a process like this.
• 38:43That's about one or two microns in diameter.
• 38:45I might go as high as two microns.
• 38:47Not much more than that because I don't
• 38:49want to get too far into the structure.
• 38:51Because when I come in here and I look
• 38:52at my data and I'm trying to say, hey,
• 38:55what is signal and what is not signal?
• 38:57What is being trained and what
• 38:58is not being trained.
• 38:59What I see on the screen here is a
• 39:02maximum intensity projection of two
• 39:05microns worth of volume in my stack.
• 39:08When I train and I come over here and say,
• 39:10you know what,
• 39:10I want to train this process right here.
• 39:12I'm going to come over here and say,
• 39:14you know what,
• 39:15I want to train this process and do this.
• 39:17I'm only selecting the pixels
• 39:18that I see on the screen.
• 39:20Even though it might be 456 different
• 39:23slices kind of merged together
• 39:24in a Max intensity projection.
• 39:27I'm only selecting the brightest pixels
• 39:30in this slice in this stack of slices.
• 39:33So they could be various levels.
• 39:35And So what I like to try to
• 39:37show people sometimes let me hit
• 39:38train predict here one more time.
• 39:39There is an idea you can use this
• 39:41to kind of show you and hide what
• 39:43the what those training voxels are.
• 39:45So if I do show hide,
• 39:46I'm going to come right to this
• 39:48guy that I just did.
• 39:49I'm going to just rotate this
• 39:50twice here to put it on its side.
• 39:51I'm going to turn it sideways
• 39:52here a little bit,
• 39:56and as I move this up and down,
• 39:57you can see the pixels that it's choosing are
• 40:00the pixels that are in that slice, right?
• 40:03That are in that section.
• 40:05This is my stroke from here to here,
• 40:07but it's selecting the brightest pixels.
• 40:10That are here.
• 40:10They happen to be a hole in a single slice.
• 40:11That wasn't a really good example.
• 40:12Let me just look at another example here.
• 40:14Here. This is a better example.
• 40:15So you can see that the stroke
• 40:17that I made was from here to here,
• 40:19all the pixels it chose,
• 40:20it chose that pixel and that pixel,
• 40:21but it didn't choose that one and that
• 40:24one because they're not visible within
• 40:26the optical slice section, right?
• 40:28And so you have this kind of
• 40:30distribution of pixels only selecting
• 40:31those brightest pixels within the
• 40:33within the view that's in that section.
• 40:35And so that's what it's selecting,
• 40:37that's what it's training
• 40:38as part of the algorithm.
• 40:39So the size matters here in terms
• 40:41of how big you want the structure.
• 40:43Again, usually you keep it about,
• 40:45I'd say the size of your structure,
• 40:47in this case about 1 1/2 to
• 40:49two microns or so.
• 40:50It's probably be good,
• 40:51but if you look at inside here,
• 40:52you can see it's doing pretty good
• 40:54job and see you can actually even
• 40:55train in this mode if you want to.
• 40:57If you wanted to come up here and say
• 40:59you know what, this is background.
• 41:00I want to cut a train.
• 41:01I can train in this this author
• 41:03slicer mode too.
• 41:04I can train at any orientation.
• 41:05I want to try to identify the structures
• 41:08because depending on the data you might say,
• 41:10oh, you know, it's not.
• 41:11I didn't do a good job training
• 41:12that bottom part,
• 41:13I'm going to come in here and train that.
• 41:14Bottom part,
• 41:14you can come in and do that.
• 41:15At any point you can come up here and say,
• 41:17you know what,
• 41:18some of this doesn't look like real signal.
• 41:20I can come up here and say you know what,
• 41:22maybe that's background,
• 41:23do that and then hit train and predict
• 41:26and kind of get that section done.
• 41:28And then again train their structures a
• 41:30little bit better within the structure.
• 41:32And I'm just, I'm just going to
• 41:33right click on this little guide,
• 41:34bring it back to normal and then you
• 41:36have your result that looks like this.
• 41:38And again as you come in here you you
• 41:40get the results that you're looking.
• 41:42Now I don't like,
• 41:43I don't think that last one made it better.
• 41:44So I'm going to go back.
• 41:45I did undo.
• 41:46There is a delete last here.
• 41:47There is no delete button or eraser button,
• 41:50but there is a delete last.
• 41:51So I like to make small.
• 41:52Once I make the original training,
• 41:54I think I make small incremental
• 41:56changes to kind of create
• 41:57this particular structure.
• 41:58But the idea here is you get a nice sharp
• 42:01border of your structure,
• 42:03much better than any of the
• 42:05other methods typically.
• 42:06And if your soma is not labeled super well,
• 42:09a lot of times cell bodies,
• 42:11their somas aren't labeled very well.
• 42:13Maybe there's a big hole with
• 42:15a nuclei is or what have you.
• 42:17With the pixel classification,
• 42:18you can train those pixels.
• 42:19Even though they're pixels within the nuclei,
• 42:21they're not very bright.
• 42:22You can train those as real pixels
• 42:24and go through and train those pixels
• 42:25and you're going to be able to get
• 42:27a nice representation of the soma.
• 42:28Even if the red labeling of the
• 42:30green labeling doesn't really
• 42:32label the nuclei very well.
• 42:33There's a big hole there.
• 42:34It's very common for neurons and
• 42:36things like that that have a very
• 42:39prominent nuclei that has very little,
• 42:41you know,
• 42:42protein standing for for
• 42:44neurotransmitters or something like that.
• 42:45And so that really helps kind
• 42:47of give you a full structure
• 42:48and volume of that structure.
• 42:50So that is the method here in terms of
• 42:52the rendering and selecting those pixels,
• 42:54whether it's with the intensity
• 42:55base or the pixel classification,
• 42:57the idea is that you're going
• 42:58to get this rendered object
• 43:00that's going to look something.
• 43:01I'm going to look at that.
• 43:02You can see the shaded area,
• 43:03it's going to look something
• 43:04along the lines of that.
• 43:05And so the idea is the next step and this
• 43:09is the step that I already processed here.
• 43:11There is a split touching object step.
• 43:13This is to help us identify
• 43:15one surface from the other.
• 43:16If those surfaces and those
• 43:18edges merge together and we can
• 43:20identify them with AC point.
• 43:22Now there's two ways to do C points.
• 43:23Now in the pixel classifier
• 43:25there is only one way and that's
• 43:27this morphological split.
• 43:28Morphological split is basically looking
• 43:30for a geometrical center of a object.
• 43:33Now in this case,
• 43:35we want to split it based on the cell
• 43:36bodies that are in this field of view.
• 43:38So the cell bodies are often
• 43:39much bigger than anything else.
• 43:41So we want this value to be relatively
• 43:43large to kind of fit a single C
• 43:45point in the geometrical center of
• 43:47these big Somas and nothing else.
• 43:49We don't want any C points in our
• 43:50processes or anything like that.
• 43:52So we set it relatively big.
• 43:53And so I open up this next step
• 43:55and step by step here,
• 43:56you can look at your data set here and
• 43:59you can adjust your threshold to try to say,
• 44:01OK,
• 44:01get one seed point for Soma.
• 44:03And it does a pretty good job in
• 44:06this particular instance to find
• 44:07because now we've made that Soma
• 44:09object with the machine learning
• 44:11that has a nice solid object,
• 44:12fully full volumetric object.
• 44:14We have a seed point there,
• 44:16C point there, C point, C point,
• 44:17those are all my cell bodies.
• 44:18Does a pretty good job.
• 44:19There might be a seed point right here.
• 44:21And I think if I lower this down a little
• 44:22bit, it'll pick that guy up.
• 44:24I put that guy right there, right.
• 44:26We try to pick up policy.
• 44:27Now, it's not always going to
• 44:28get every single one right,
• 44:29and there's a lot of other
• 44:30processes here that are not
• 44:31really attached to the structure.
• 44:32But what this process does,
• 44:34what the C point thresholding does,
• 44:36is it helps us split these objects
• 44:38that might be touching each other.
• 44:39So you get something that looks
• 44:41like this and you can see all
• 44:42of a sudden now we get a surface
• 44:43rendering that looks like that.
• 44:44We got a pretty good representation
• 44:46of these Somers.
• 44:47You have the processes,
• 44:47you have a lot of different
• 44:49structures in here. Again,
• 44:50I'm going to just finish this process here.
• 44:52And all those little
• 44:53processes that were in here,
• 44:55they were just disconnected processes,
• 44:56probably part of another Microglia
• 44:58that wasn't in this section, right.
• 45:00It's not connected, it's not,
• 45:02it's not continuous with the
• 45:03structures that we have here.
• 45:04Now again,
• 45:05this is a really good example and
• 45:07the reason this works well with
• 45:08Microglia is that they're often
• 45:10nice and continuous structure.
• 45:11So if I view this here with object ID,
• 45:13you can kind of get a better
• 45:15idea of the structures here.
• 45:16That's a microglia,
• 45:17single microglia, microglia,
• 45:18single microglia.
• 45:19You have these structures now with these
• 45:21individual structures that we have that
• 45:23were created using the pixel classification,
• 45:26using the splitting to try to identify
• 45:28these microglia as best we can.
• 45:30And again,
• 45:30if it's continuous with this structure,
• 45:33it's going to be a single microglia.
• 45:34Now this one might go a little bit too far.
• 45:36I don't know if it's connected a little
• 45:37bit further than the word should,
• 45:38is a little bigger than everything,
• 45:39but maybe that's a real microglia.
• 45:41Again, if it's continued with the
• 45:42structure that's what we're rendering.
• 45:44And so if it's not continuous then
• 45:46we're we're it's not going to kind of,
• 45:47it's not going to split it and
• 45:49it's going to keep it,
• 45:49keep it separate.
• 45:50So that is kind of surface
• 45:52rendering in a nutshell,
• 45:54all the different ways we had to
• 45:55make surfaces within the structure.
• 45:59The next step is to kind of look
• 46:01at say look at some statistics.
• 46:02So if we were to take this data set
• 46:04and for example the microglia and say,
• 46:06hey, what kind of statistics
• 46:07do we have from the structure,
• 46:08we have the idea to kind of come
• 46:10in here and look at the statistics.
• 46:12So if you click on the statistics tab,
• 46:14here you go to detail,
• 46:15all these values are listed here.
• 46:16For your statistics,
• 46:18we have surface area, we have volume,
• 46:21we have intensity.
• 46:23All these values are here that we
• 46:26can measure for each one of those
• 46:27and they're all selected here.
• 46:28And if you click on the gear tab
• 46:30here or the the disk here will
• 46:32export what you have on display here.
• 46:34Or if you click this guy here that'll
• 46:36export everything as a CSV file and
• 46:38you can open that up in excel and
• 46:40export all the statistics together and
• 46:41and kind of deal with all the stats
• 46:43that you want to get within the structure.
• 46:45Now one statistic that I like a lot and
• 46:47I use a lot for data sets like this
• 46:50is I use bounding box feature here
• 46:53which gives you the size of the structure.
• 46:54So it every surface in a
• 46:56Morris is contained by a box,
• 46:58a 3D box.
• 46:59This bounding box I finds the longest
• 47:01principal access of the of the of the cell.
• 47:04So for example,
• 47:05this blue guy that's in here probably
• 47:07has a really long bounding box 108,
• 47:09which is measuring the edge of that
• 47:11surface from there,
• 47:12probably the edge of that surface there
• 47:14along the the longest principal axis
• 47:16of a box that fits that structure.
• 47:18And then some of these smaller
• 47:19guys that are that are here,
• 47:20they're going to be like 63,
• 47:21it's going to be much smaller.
• 47:23Now that's not a complete structure,
• 47:24but it's edge of the structure.
• 47:25The edge of the structure
• 47:26gets that bounding box,
• 47:28gives you a little bit of
• 47:29morphological measures out there.
• 47:30But any of these statistics you can
• 47:32export and get out of the structure.
• 47:34Now I'm going to go through here really
• 47:36quickly just to give you a concept
• 47:38of some analysis that you can do.
• 47:40I'm not going to make the surface,
• 47:41I'm not just going to walk through
• 47:42it because I've already done it.
• 47:43You can do it with the traditional
• 47:44means here or you can do the
• 47:46pixel classification means I did
• 47:48both Results are very similar,
• 47:49but the idea is you don't always
• 47:51need the pixel classification,
• 47:52just depends on what you're looking for,
• 47:54on how you're trying to render your objects.
• 47:55Sometimes for puncta the the the
• 47:59the regular creation parameters
• 48:00within the surface works works
• 48:02really well and you can see we've
• 48:04made all these little punks.
• 48:05And again, if we go into the Slicer mode,
• 48:07turn this off,
• 48:08make this white,
• 48:14right, you can see these
• 48:16surfaces are getting rendered.
• 48:17Any surfaces again that are touching,
• 48:19that's a punk.
• 48:20And that's a puncta, right?
• 48:21AC point there and AC point there,
• 48:23they're going to get split.
• 48:24So you have a surface there
• 48:24and you have a surface there,
• 48:25two separate surfaces based on the
• 48:27C point splitting in that watershed
• 48:29that we talked about earlier, right.
• 48:31So then we have the numbers, right?
• 48:32You can come up here and you can
• 48:34say how many Puncta do we have?
• 48:35How many punk did we create?
• 48:37Click on here go to Overall,
• 48:38we have 3861 total punked up in
• 48:42this entire volume, top to bottom.
• 48:44And again,
• 48:44if we turn this off and look at it,
• 48:46that's what it looks like.
• 48:46Little tiny little surfaces
• 48:48pockmark throughout the data set.
• 48:51Now, one of the big things
• 48:53that a Mars is known for,
• 48:55and what most people use with a
• 48:56Mars is they want to look at the
• 48:58interaction between multiple surfaces,
• 48:59whether they're surfaces or
• 49:01spots or surfaces and surfaces.
• 49:02We do those sorts of things.
• 49:05So just to kind of before I get into that,
• 49:07I'm sorry, before you get into that,
• 49:08you can do the spot detection as well.
• 49:10I'm not going to spend a whole lot
• 49:11of time on the spot detection,
• 49:12but I just want to kind of show you,
• 49:13show you what that does here real quick here.
• 49:15So a lot of times people just
• 49:17want to count spots.
• 49:18They don't,
• 49:19they don't really care about the morphology,
• 49:21they don't care about the size.
• 49:22They just want to know how many
• 49:23punked are in my field of view, right.
• 49:25That's where the spot feature
• 49:27comes into play.
• 49:28And again,
• 49:28if I come in here and I look at those spots,
• 49:30let me turn this off.
• 49:34You have these little spots
• 49:35within the data set.
• 49:36Again,
• 49:36you can view this in a slicer mode
• 49:38and you'll see a little circle and you
• 49:39can see those spots identifying all
• 49:41the different puncta within the data set.
• 49:43And again,
• 49:43if I come up here and rebuild it,
• 49:45this is a really simple feature.
• 49:48You come in here, you choose a spot size,
• 49:50single spot size, 1 Micron, 2 microns,
• 49:53whatever the size you're trying to do,
• 49:55and identify these functa.
• 49:56Typically you do a little
• 49:57background subtraction as well.
• 49:58You don't have to.
• 49:58If you want to just do it on the raw data,
• 50:00you can do that as well.
• 50:01But background subtraction,
• 50:01again on a data set like this that has a
• 50:04fair bit of signal noise in your tissue,
• 50:05it's probably a good idea in most cases.
• 50:07I almost always have that turned on.
• 50:09You come in here and you're
• 50:10going to get a threshold.
• 50:11You adjust the threshold to
• 50:13make sure your spots are real.
• 50:16Identify a particular threshold,
• 50:17you hit finish and then you can get all your
• 50:19spots that are created in that data set.
• 50:21You can see all your spots in that data set.
• 50:23And again, if you come in here,
• 50:24your statistics,
• 50:24you have the total number of
• 50:26spots within your structure.
• 50:27Really quite easy to get those
• 50:29numbers within the structure.
• 50:30But the big thing is how do we do
• 50:33the analysis of this CD 86 surface
• 50:36here and this Microglia one here?
• 50:39I'm going to do,
• 50:39I'm just going to do this one here.
• 50:40That's really right.
• 50:43So we have these microglia that we rendered,
• 50:45right.
• 50:45So those are our, our microglia surfaces.
• 50:47We want to maybe measure the the
• 50:51relationship between the two.
• 50:52So the biggest thing that we like to
• 50:54do is we're going to come in here and
• 50:57turn on this object, object statistics.
• 51:00So under the edit tab.
• 51:01Now I didn't turn it on for all
• 51:02of them because I don't want to
• 51:03turn the law for all of them.
• 51:04I only only have these turned on.
• 51:05I only have it turned on for
• 51:07these two guys right now, right.
• 51:08So I have you don't want to turn on
• 51:09for everything because then it's going
• 51:10to calculate everything and that can really,
• 51:12depending on the size of your
• 51:13data set that can really bog
• 51:14down your system a little bit.
• 51:15So we try to only turn it on when
• 51:16we're ready to kind of calculate
• 51:18and export a particular statistic.
• 51:20And So what I want to know is
• 51:21what the customer want to know.
• 51:23This is a real customer data
• 51:24set from a few years ago.
• 51:25They wanted to know how much of the
• 51:28CD 80 AC-6 protein is overlapping
• 51:31with the microglia surface that
• 51:32we've rendered in its entirety,
• 51:34hopefully right with these
• 51:36surface renderings, right.
• 51:37So we made the surface renderings
• 51:38of the microglia,
• 51:39we made the surface renderings of the ACD
• 51:4186 making these volumetric measurements,
• 51:43and they wanted to know how much
• 51:45of the CD 86 is overlapping, right?
• 51:46And so by turning on this object to
• 51:48object statistic makes it really,
• 51:50really easy because what happens
• 51:52now if I click on IBA one and
• 51:55I go to my statistics tab,
• 51:57put a detail here,
• 51:59I'm going to Scroll down here,
• 52:00there's a value here that says overlap
• 52:03volume ratio with CD86, right?
• 52:05So CD 86 is the the name of the the
• 52:08the surface structure we created here.
• 52:10So I'm going to click this
• 52:11Overlap volume ratio.
• 52:12Overlap volume ratio is basically a
• 52:15percentage of how much the green surface
• 52:17is overlapping with the red surface, right?
• 52:19And so every red surface.
• 52:21So this guy has a red surface
• 52:23and that's the number .06.
• 52:26So that surface that re rendered
• 52:28in the red is overlapping 0 point.
• 52:31The ratio is .06.
• 52:32But if you multiply that by 100,
• 52:34it's basically 6% is what it's saying.
• 52:36So 6% of the volume,
• 52:39total volume of that microglia is
• 52:42overlapping with green pumped up, right.
• 52:45So about 6%.
• 52:46This guy here is 8%, this guy here 5%,
• 52:50this guy here 7%.
• 52:51So you can get a range of distribution
• 52:53to say how much of the microglia
• 52:55is actually overlapping with this
• 52:57activation protein that is supposed to
• 52:59be kind of on the inside of that structure.
• 53:02And you have this way of kind of getting
• 53:03these interactions to meet to each other.
• 53:05You could also look at distances,
• 53:07So this is looking at the overlap ratio.
• 53:10You can also look at distances.
• 53:11So if I wanted to go back to the
• 53:14CD 86 and say you know what,
• 53:16I wanted to find the CD 86,
• 53:17not just just the ones that are overlapping.
• 53:19I want to get the CD 86 that's
• 53:22within X microns of my micro glare
• 53:24or something along those lines.
• 53:25Right. I can come down here.
• 53:28There's a filter tab here.
• 53:29I'm going to click a filter so
• 53:31this is the filter tab here and you
• 53:32can use this at any point here.
• 53:34Scroll down here.
• 53:35That statistic here say shortest
• 53:37distance here,
• 53:38So we're going to do shortest
• 53:39distance to IBA one, right?
• 53:41Which is the surface that we've created,
• 53:43the intensity based.
• 53:44And I can set this threshold and say,
• 53:46hey, you know what I want Any punk
• 53:48that outs within two microns,
• 53:49there we go, right.
• 53:51So out of 300 and six 3861,
• 53:54we have 21110 that are within two microns.
• 53:58And I can duplicate this out and
• 53:59I can get a new structure here.
• 54:01And I usually like to rename this
• 54:02because it's really usually a
• 54:04long rename here.
• 54:04So I'll come up in here and I'll
• 54:06do this C86 within two microns
• 54:11of IBA one whatever or IBA one
• 54:17with an IBA one and you get quantification.
• 54:19Here you can look at the statistics.
• 54:24Again, I like to turn when I do a
• 54:26duplication, I like to turn this off.
• 54:27But again, if you go to your statistics here,
• 54:29go to overall, you have your number
• 54:31right here and you can export
• 54:33that number out using this button.
• 54:34You can write it down.
• 54:35You can do a lot of different stuff
• 54:36to kind of get that numbers out there,
• 54:37but you can get some numbers
• 54:39there from the structure.
• 54:40And again, you can turn on your structure
• 54:43here and identify those structures.
• 54:45Now visually, you might want to
• 54:48make this a little bit prettier.
• 54:49Switch this off here.
• 54:51Say you wanted to turn this on to base color.
• 54:55Make it red.
• 54:57I can make it transparent.
• 55:00My
• 55:12colors are weird. My colors? Weird.
• 55:22Black. Interesting. All right, Black second
• 55:29here. Oh, I'm not sure why they're black. Let
• 55:36me just save this real quick.
• 55:37When I'm going to reopen the file,
• 55:40something's wrong with my graphics
• 55:41card here. They shouldn't be black,
• 55:45it's just re saving the image here.
• 55:54Yeah,
• 55:59so the idea is that you
• 56:01can make a little better.
• 56:02Yeah instead of graphics glitch there,
• 56:05you can make these a little bit transparent.
• 56:07So for example if I want to show
• 56:09just the ones that are overlapping,
• 56:11there they are, right?
• 56:12And if I zoom in, you can make a movie
• 56:16and you can actually show them which
• 56:18ones are inside of the structure.
• 56:19And you can say, since I did,
• 56:21I did like 2 microns.
• 56:22Some of them are going to be a
• 56:23little bit outside the structure.
• 56:24But again, you can set that
• 56:25threshold to whatever you want.
• 56:26You can set the ones that are overlapping,
• 56:27only the ones that are totally inside.
• 56:30There's a lot of different features here,
• 56:31but the idea is that you can
• 56:33really render these guys kind of
• 56:35interesting ways to kind of visualize
• 56:37structures inside structures.
• 56:38You can do the same type of analysis with
• 56:41this box as well and you can do that as well.
• 56:44So what are we at 10:30 here?
• 56:45OK, so I'm going to move on
• 56:49here from surfaces.
• 56:50I'm going to jump very, very,
• 56:51very quickly just to kind of
• 56:52give you a quick preview.
• 56:54There's a couple of really great
• 56:55tutorials that I can I can point
• 56:57you to moving forward through our
• 56:58in our Learning Center that I made
• 57:00a few years ago or just last year,
• 57:02a year and a half ago about filament tracer.
• 57:05So filament tracer is a feature
• 57:07of a Morris where especially for
• 57:09data sets like this.
• 57:10And again,
• 57:11I don't know exactly what
• 57:13your strong suits are,
• 57:14the core facility,
• 57:15so there's a lot of different features and
• 57:17and images that people are going to get.
• 57:19But film and tracing is a tool that has
• 57:22a lot of different options to be able to use.
• 57:25In this particular case,
• 57:26someone might want to try to
• 57:28measure the number of branches,
• 57:29the terminal points,
• 57:31size of the Somas and and things like that.
• 57:34Again,
• 57:34I'm not going to walk
• 57:35through the wizard here,
• 57:36I'm just going to show you kind of
• 57:37pointing fact kind of the end result
• 57:40where you can identify as long as
• 57:41you have a nice starting point.
• 57:43In this case we do nice little soma
• 57:45we can come up here and we can create
• 57:47something that looks like that.
• 57:49So this is not going to be exactly
• 57:52the same as the surface rendering.
• 57:54It might look a little bit different
• 57:56depending on where your seat points are
• 57:58in the training and things like that.
• 57:59But the idea is as long as you do
• 58:03even nice clean data set and you can
• 58:04kind of clean the files up to kind
• 58:06of get where their structures are,
• 58:07we get a statistic here,
• 58:10click on the filament object.
• 58:11We have a filament that looks like that.
• 58:13And so from this particular
• 58:15structure we can get detailed
• 58:17information about that filament.
• 58:18We can get SOMA size.
• 58:21So we do have now Soma that we've identified.
• 58:23You can see the Soma as
• 58:24blue dots in the middle.
• 58:26So you can come up here.
• 58:26If you do it right and you have a
• 58:28nice data set, you can get SOMA,
• 58:31there's volume, soma volume, right.
• 58:33So there's the soma volume.
• 58:35That's the structure of that structure there.
• 58:37And getting the size of that soma,
• 58:39Soma volume is sometimes something
• 58:40people really want in terms of
• 58:42morphology and things like that.
• 58:44Most people are looking for things like
• 58:46branch level, number of branch points,
• 58:48you know, number of branch points.
• 58:49There we go. This guy here has four,
• 58:51you know, 14 branch points.
• 58:53It's tiny. This guy here,
• 58:56110 branch points, right?
• 58:57We can kind of identify the complexity
• 58:59of the structure.
• 59:00Now again, the cleaner the data,
• 59:02the better the structure,
• 59:03the more accurate this trace is going to be.
• 59:04But this was done all 100%
• 59:06automatically identifying the soma,
• 59:08finding the C points and it
• 59:09traces this path as best you can.
• 59:11And so you can apply this to multiple
• 59:13images and get a good idea of the the
• 59:15complexity of some of these neurons,
• 59:17whether it's number of branch points,
• 59:19number of terminal points.
• 59:23There's a Shoal analysis in here as well.
• 59:25If you're interested in Shoal
• 59:26analysis through Shoal intersections,
• 59:27you get the number Shoal intersections for
• 59:29each one of these filaments on their own.
• 59:30It's a lot of different ways to kind
• 59:32of measure dendrite complexity.
• 59:34Like I said, if you have some
• 59:35data sets that are like this,
• 59:36this would be a whole hour and a half
• 59:39session on its own just to talk about
• 59:40how to do film and tracer and the
• 59:41different methods there to do that.
• 59:43If you have any questions on that,
• 59:44please just reach out to the support team,
• 59:46send me some files.
• 59:47We can walk through it together to do
• 59:48a little bit of more in depth training
• 59:49or you can watch some of the videos
• 59:51that are on on our website to kind of
• 59:54learn a little bit about how that works.
• 59:57One of the things I also wanted
• 59:58to kind of bring up today that I
• 01:00:00think is really important for kind
• 01:00:02of new users to kind of know and
• 01:00:03understand what it's doing and why
• 01:00:05it's there because I kind of glanced
• 01:00:07through it in the wizard process.
• 01:00:09So I'm going to switch gears a little bit.
• 01:00:11I'm going to get away from
• 01:00:12this particular data set.
• 01:00:13I'm going to move to a 2D data set.
• 01:00:16Now, 2D data sets work just
• 01:00:18as well as as 3D data sets,
• 01:00:20but the idea here is I wanted to
• 01:00:22show you a feature that I think is
• 01:00:23really interesting and it can be used
• 01:00:25across the board, whether it's 2D,
• 01:00:27whether it's 3D, whether it's whatever.
• 01:00:28And this is a data set.
• 01:00:30This here is a brain slice.
• 01:00:33Again, it's just an example.
• 01:00:35I know, I don't know if you're all
• 01:00:36neuroscientist or not, but I am.
• 01:00:37So I get really excited when I see
• 01:00:39brain and neurons and things like that.
• 01:00:41But in this particular example,
• 01:00:43this is a single brain slice.
• 01:00:46You can see the tissue,
• 01:00:47the curve of the brain Here,
• 01:00:49I think it's a cortex or something.
• 01:00:51Exactly.
• 01:00:51Sure.
• 01:00:52Hippocampus, actually.
• 01:00:53And so the idea is the the user wants to be
• 01:00:58able to count these cells Now in this image,
• 01:01:01whether it's red,
• 01:01:02green or blue,
• 01:01:02whatever channel you're looking in,
• 01:01:05it's not just labeling cell bodies, right.
• 01:01:07So if you look at this channel,
• 01:01:08it's labeling a lot of other stuff.
• 01:01:11So that makes it really
• 01:01:14difficult to render cells.
• 01:01:16Now there's other cells in here too.
• 01:01:19Then this is a brain slice.
• 01:01:20There's a thousands of other cells in here.
• 01:01:21So if you did Dappy or something like that,
• 01:01:23you would have Dappy cells and neurons
• 01:01:25everywhere out throughout the data set.
• 01:01:27So we want to try to render these
• 01:01:29guys with a spot and then maybe
• 01:01:30they want to count the cells,
• 01:01:31maybe they want to look
• 01:01:33for colloquialization.
• 01:01:33In this particular case,
• 01:01:34they wanted to look for triple cologue cells.
• 01:01:36The CFVCFPYFP and these DS red
• 01:01:39cells label three different markers.
• 01:01:40They wanted to say, hey,
• 01:01:41which ones are labeled with what.
• 01:01:43Now the challenge is when you do,
• 01:01:46for example, spot detection
• 01:01:47for the green cells,
• 01:01:49what you get is something that looks like
• 01:01:52that this is the raw quantification, right?
• 01:01:54You get spots everywhere, right?
• 01:01:57We're we're identifying the spots here.
• 01:01:59We're also getting all these spots along
• 01:02:01these axons and all these spots here,
• 01:02:02but all the cells are pretty
• 01:02:04detected pretty well.
• 01:02:05That's a single slice because some of them
• 01:02:06are going to be a little bit out of focus.
• 01:02:08But for the most part, again,
• 01:02:09you're going to have to take that for what
• 01:02:11it is because it is just a single slice.
• 01:02:12But the idea is we're trying to
• 01:02:14identify all the cells that are in
• 01:02:16focus in this field and quantify them.
• 01:02:19But part of the problem is as you lower the
• 01:02:20threshold to detect some of the dimmer cells,
• 01:02:22even even some of the brighter cells,
• 01:02:24you get a lot of false positive
• 01:02:27counts with this particular process.
• 01:02:29Now with the Morris, we have a great
• 01:02:32tool that's built into the wizard.
• 01:02:35Now I didn't.
• 01:02:35I'm going to just rebuild it
• 01:02:37here from the start here.
• 01:02:38So it's called classification.
• 01:02:39So when you start a new spots object
• 01:02:41or surface object for that matter,
• 01:02:43there's this classify objects spots again,
• 01:02:46often I start with it turned
• 01:02:47off and then if I need it,
• 01:02:48I go back and I turn it on.
• 01:02:49If I know I'm going to use it,
• 01:02:50then I'll turn it on.
• 01:02:51But basically it's a label feature where
• 01:02:54you can have it learn through machine
• 01:02:56learning and and standard filtering as well.
• 01:02:59Have it label,
• 01:03:00identify and classify one object
• 01:03:02into into one category,
• 01:03:04another object in another category
• 01:03:06based on the the the structures
• 01:03:09that you that you segmented.
• 01:03:10This is the last step in the one of
• 01:03:12the last steps in the segmentation.
• 01:03:14After you've created the segmentation,
• 01:03:15whether you do pixel classification,
• 01:03:16spot detection, regular creation,
• 01:03:18doesn't really matter.
• 01:03:19This last step is going to take
• 01:03:21that whatever you rendered,
• 01:03:22whatever you detected and try to classify
• 01:03:24them based on the training that you give it.
• 01:03:27And so this is another level of
• 01:03:28machine learning instead of kind of
• 01:03:29pixel based machine learning where
• 01:03:30you're kind of using it to kind of
• 01:03:32select the pixels and make the surface.
• 01:03:33This is more of an object based machine
• 01:03:35learning which the objects already generated.
• 01:03:38There are statistics related to that object,
• 01:03:40various different statistics and
• 01:03:41it's going to try to categorize
• 01:03:43them based on those statistics,
• 01:03:45lot of different statistics.
• 01:03:46And so I'm going to kind of jump through
• 01:03:48here again, I'm not going to spend,
• 01:03:51OK, I'm going to do this, sorry,
• 01:03:53I'm going to delete this here,
• 01:03:55I'm going to go,
• 01:03:56I'm just going to rebuild this guy here,
• 01:03:57right.
• 01:03:58So we do classify spots.
• 01:03:59I did A7 Micron spot size.
• 01:04:02You come in here and you get a whole bunch
• 01:04:04of spots right?
• 01:04:04So I set this to 184 right?
• 01:04:07184 is fine. If I set it too low here,
• 01:04:11you might miss some spots,
• 01:04:13might miss themselves,
• 01:04:14maybe they're not quite in focus,
• 01:04:16maybe you don't want to include them.
• 01:04:17But even if you did include them,
• 01:04:19there's still some spots out
• 01:04:20here that you don't want.
• 01:04:21But the idea is that I want to try
• 01:04:23to count all the spots that I can
• 01:04:24that are in this field of view,
• 01:04:26even those dim guys that are there.
• 01:04:28So I lower this threshold a little bit more.
• 01:04:30And when I when I set it really low,
• 01:04:31I get a lot of background and noise.
• 01:04:34So I set this to like 184
• 01:04:36or whatever it was here,
• 01:04:38but you get a lot of different spots here.
• 01:04:39And then when I go to the next step,
• 01:04:43you can see I've already done the
• 01:04:45training here in this particular example.
• 01:04:46But the idea is that you have this
• 01:04:48training data set under here,
• 01:04:50there's a a classification step,
• 01:04:52you create a new classification,
• 01:04:53hit this drop down menu,
• 01:04:54come down here to machine learning.
• 01:04:56And the idea is pretty simple.
• 01:04:57You're going to train which ones
• 01:04:59are real and which ones are not.
• 01:05:01So I'm going to, I'm going to keep this one,
• 01:05:03but I'm going to just create a brand new one
• 01:05:05here just to show how it works here, right.
• 01:05:06So if I come in here and I say,
• 01:05:08you know what?
• 01:05:09I know all these guys are garbage.
• 01:05:10So I can hold down control and
• 01:05:11I can say I can come down here,
• 01:05:13I know those are all garbage.
• 01:05:14I can select,
• 01:05:14I can select a lot of these really easily.
• 01:05:17These are all garbage.
• 01:05:17And I want to select as many
• 01:05:19as I possibly can.
• 01:05:20Not only ones that are down
• 01:05:21here that are garbage,
• 01:05:22but also ones that are up here,
• 01:05:24that are garbage, that have dimmer
• 01:05:25backgrounds and things like that.
• 01:05:27I'm going to put those into Class
• 01:05:29B 103 trained cells, right?
• 01:05:31And I'm going to zoom in a little bit closer.
• 01:05:33Now I'm going to start training
• 01:05:34ones that are real,
• 01:05:35and I can hold down the control
• 01:05:36button as well.
• 01:05:36Select these guys that are real,
• 01:05:38select all the guys that I think
• 01:05:40are real cells that I want to keep.
• 01:05:42Put those into Class A,
• 01:05:43right?
• 01:05:43And you're going to go through
• 01:05:45and you're going to train as many
• 01:05:46of these as you possibly can.
• 01:05:48The nice thing about this machine
• 01:05:50learning in any of the machine learning,
• 01:05:51I didn't talk about this before,
• 01:05:52but any of the machine learning
• 01:05:54and the surface creations that
• 01:05:55we've created in the earlier part
• 01:05:57of the presentation,
• 01:05:58you can save those parameters and
• 01:06:00you can apply them to other images.
• 01:06:02So for example,
• 01:06:03that machine learning that I
• 01:06:04did on that Microglia data set,
• 01:06:06I can save those parameters from
• 01:06:08that training and I can apply them
• 01:06:10to an identical a similar data set
• 01:06:11that was required in the same way,
• 01:06:13same quality,
• 01:06:14and I can apply the exact same pixel
• 01:06:16classification to those particular data sets.
• 01:06:18Same thing here.
• 01:06:19As soon as I train these data sets
• 01:06:21and I say you know what these
• 01:06:23are all these are all the cells
• 01:06:24that I want to identify.
• 01:06:28When I save this spots current parameter,
• 01:06:31this training that I give will be identified.
• 01:06:34So all these guys here on those lines,
• 01:06:35those are all background.
• 01:06:36Make sure we choose those background.
• 01:06:38These guys are background and again you're
• 01:06:41going to try to get as many of these
• 01:06:43guys as you possibly can that are real,
• 01:06:45these guys that are not real.
• 01:06:46And again, usually I'm going to train
• 01:06:48a little bit more than this year,
• 01:06:49but the idea is that hit train and predict,
• 01:06:52you're going to start seeing all
• 01:06:54those guys are going to turn red and
• 01:06:55those guys are going to turn green.
• 01:06:57Now this isn't a once and done deal, right?
• 01:06:58You can say, hey, you know what,
• 01:07:00that's a cell.
• 01:07:01I want to keep that guy that's a cell,
• 01:07:03put it in that, put it in that category,
• 01:07:04train and predict again.
• 01:07:05And it's going to hopefully chain
• 01:07:07chain the categories elsewhere.
• 01:07:09And again,
• 01:07:09you just kind of look through your
• 01:07:11data set and do the best training
• 01:07:12that you possibly can to identify
• 01:07:14what is a cell and what's not a cell.
• 01:07:16Again, the more you train,
• 01:07:17the better it's going to bet.
• 01:07:18And you can see it's doing a pretty
• 01:07:20pretty decent job of identifying find
• 01:07:22these guys that are real cell bodies that
• 01:07:24are in focus that I want to include.
• 01:07:27And so the idea is that now I have these
• 01:07:29training data sets and hit finish.
• 01:07:30When I get the labels here I can
• 01:07:32see my labels there are set.
• 01:07:34This is set number two of the ones I
• 01:07:36did here I have 215, good ones 12151.
• 01:07:42And So what we ended up doing in this
• 01:07:44particular instance is I did it for all
• 01:07:46three Red, green and blue come down here.
• 01:07:49And the idea is that I made a separate
• 01:07:51group here of the classifications.
• 01:07:53So these are my red, my Blues,
• 01:07:55my Reds, my green.
• 01:07:56And then I did that.
• 01:07:57I just did the triple Colo.
• 01:07:59So here's here are triple Colo green cells.
• 01:08:01So if I go back here and
• 01:08:03I look at these guys,
• 01:08:04these guys here are triple Colo green cells,
• 01:08:08that guy.
• 01:08:08And if I turn on my red and
• 01:08:11I turn on my my blue, right,
• 01:08:14You can see that there's a spot
• 01:08:15on each one of them, right?
• 01:08:17That's a triple colloquized.
• 01:08:18You have double colloquized
• 01:08:20ones here as well.
• 01:08:21You have single red cells.
• 01:08:22You have single blue cells.
• 01:08:24You have single green cells.
• 01:08:25You have all these structures here
• 01:08:27you have for your segmentation,
• 01:08:29you have the the way of identifying
• 01:08:31whether or not they're they're like,
• 01:08:34here's there's blue and pink or blue and red,
• 01:08:37blue and red double labeled cells,
• 01:08:38not green labeled, right.
• 01:08:39There's another triple labeled cell.
• 01:08:41And so there's a lot of
• 01:08:42different ways to do it.
• 01:08:43But that machine learning is a really,
• 01:08:44really nice trick to help remove
• 01:08:48kind of incorrectly placed spots.
• 01:08:51Or surfaces for that matter,
• 01:08:52does the same thing with surfaces.
• 01:08:54This happens maybe if you have
• 01:08:55a lot of noise or background
• 01:08:56and and things like that.
• 01:09:00OK, so gosh, what time is it
• 01:09:0710:45? OK, so
• 01:09:11I'm getting to the end of kind of the
• 01:09:13main things that I wanted to cover today,
• 01:09:16some of the things that we didn't cover
• 01:09:18that are still of interest to a lot
• 01:09:21of people is tracking, self tracking.
• 01:09:24If you have a time lapse data set,
• 01:09:27all of the surface rendering that
• 01:09:28we've done so far, the spot creation,
• 01:09:30the surface creation,
• 01:09:31everything that we've done can be
• 01:09:33applied in a time lapse and used
• 01:09:35our tracking algorithm to that
• 01:09:37track those objects over time.
• 01:09:39So the better you can, again it goes,
• 01:09:41always goes back to acquisition time,
• 01:09:43resolution, those sorts of things.
• 01:09:44The better you can acquire your data,
• 01:09:46the better you can render individual cells,
• 01:09:49whether it be a spot object
• 01:09:50or surface object,
• 01:09:51the better the tracking algorithm
• 01:09:52is going to work as well.
• 01:09:54And again, tracking just real
• 01:09:55quick what that looks like here.
• 01:09:57If I open up a demo data set,
• 01:10:00let me see if I have one,
• 01:10:07I can do this one. Sure.
• 01:10:08This is a little bit of a unique data set
• 01:10:11where I actually use the machine learning
• 01:10:12to clean up the data set a little bit.
• 01:10:14So for example, the raw data
• 01:10:18looks like this, right?
• 01:10:20So this is not a very clean data set.
• 01:10:22The customer allowed me to share
• 01:10:24this data set because I thought it
• 01:10:25was kind of interesting.
• 01:10:26But this is I think a blood vessel and
• 01:10:29there's cells going through this vessel here.
• 01:10:32And so in time, you can kind of see these
• 01:10:35guys kind of moving along in space.
• 01:10:37Now the time resolution here isn't fantastic,
• 01:10:39but you can see there's a lot of signal
• 01:10:41in there that are clearly signal,
• 01:10:43but there's also a lot of garbage and
• 01:10:45other stuff floating around in here.
• 01:10:47So what I ended up doing in this
• 01:10:49particular instance is I was able to
• 01:10:51kind of detect all the cells here.
• 01:10:53I did the blue cells with the machine
• 01:10:56learning and identified the cells as
• 01:10:57well as a lot of different background
• 01:10:59and noise structures as well.
• 01:11:00But with a little bit of machine learning,
• 01:11:03it was able to detect most of not all of,
• 01:11:05but most of the cells that are moving
• 01:11:07through the vessel and all the ones that
• 01:11:08are on the edge of the vessels here,
• 01:11:10these pink guys,
• 01:11:11it was able to kind of identify those
• 01:11:13as not really objects of interest.
• 01:11:15And so I was able to kind of
• 01:11:17isolate those blue cells blue,
• 01:11:19duplicate them out,
• 01:11:20get them outside by themselves
• 01:11:22here and then build the tracking
• 01:11:24algorithm just on those blue cells.
• 01:11:26Now I'm not detecting them separately,
• 01:11:27so I just go,
• 01:11:28I jump right into the wizard here and
• 01:11:30jump right to the tracking step, right.
• 01:11:31And in the tracking step,
• 01:11:33you have a couple different options here.
• 01:11:34There's some autoregressive motion,
• 01:11:36Browning motion.
• 01:11:37Those are the two big ones in terms
• 01:11:39of movement algorithms that we have
• 01:11:41to kind of trace objects over time.
• 01:11:43There is a lineage algorithm as well.
• 01:11:45This looks for branching.
• 01:11:46So if you were doing cell division,
• 01:11:48Embryology,
• 01:11:48those sorts of things,
• 01:11:49and you had a decent enough time resolution
• 01:11:52to kind of watch cells divide and try
• 01:11:54to find their pattern or their generations,
• 01:11:56then this lineage algorithm will try to,
• 01:11:58if it's present,
• 01:11:59and if it can try to detect
• 01:12:00it based on the algorithm,
• 01:12:01it'll try to detect when
• 01:12:03there's a splitting cell,
• 01:12:04a dividing cell into two.
• 01:12:06And it works pretty good.
• 01:12:07Again, as long as you have good
• 01:12:09enough clear enough signal,
• 01:12:10good enough time resolution so that you
• 01:12:12can actually see it actually splitting,
• 01:12:13it works really,
• 01:12:14really well.
• 01:12:15But in this case,
• 01:12:16we're going to use probably autoregressive
• 01:12:18motion which kind of looks for kind of
• 01:12:20predictive patterns and the pathway,
• 01:12:21these fibers,
• 01:12:22these cells rather moving in One
• 01:12:24Direction from left to right,
• 01:12:26you know,
• 01:12:26and you can,
• 01:12:27you can track these
• 01:12:28objects and you can see the
• 01:12:29tracks of some of the objects here.
• 01:12:30And again, it's not tracking him 100%
• 01:12:33perfectly, but it's tracking a lot of
• 01:12:34these guys fine for a couple different
• 01:12:36time points until it kind of loses it.
• 01:12:38Like this guy right here,
• 01:12:39if I go up and back and forth,
• 01:12:40you can see it's tracking
• 01:12:41that guy pretty nicely.
• 01:12:42So it kind of appears out of nowhere,
• 01:12:44tracks it, tracks it,
• 01:12:45tracks it, tracks it, tracks it.
• 01:12:47And then maybe it events where he
• 01:12:49kind of loses it because now it's not
• 01:12:51100% sure if that's the same guy.
• 01:12:52So it stops that track and then it moves
• 01:12:54on to get another track and track these guys.
• 01:12:56And so from from the data here that
• 01:12:58you get from from tracking these guys,
• 01:13:01you have a whole slew of new statistics spot.
• 01:13:05Oops.
• 01:13:05Oh, that track crashed on me Live demo.
• 01:13:10We have a lot of different
• 01:13:13ways of exporting that data,
• 01:13:17different tracking statistics,
• 01:13:20instantaneous speeds.
• 01:13:21You have individual object tracks here.
• 01:13:25Let's put this back up here in here
• 01:13:32again, whole different set of statistics.
• 01:13:33But now you have some tracking data.
• 01:13:35So if I look at, for example,
• 01:13:40let's just look at speed, right.
• 01:13:43So if I check an object that I'm tracking
• 01:13:45that has speed check on this guy here.
• 01:13:47Now a lot of these guys aren't
• 01:13:48tracks super duper long,
• 01:13:49so the tracks are really small,
• 01:13:50but you can get instantaneous tracking.
• 01:13:53You know, just looking at the graph here,
• 01:13:54this orange is representing the track
• 01:13:56that I have selected right here.
• 01:13:57I can select another guy
• 01:13:59here that's there too.
• 01:14:00But the idea is that you can
• 01:14:01get that tracking data out say,
• 01:14:03hey, that that's the speed at
• 01:14:04that particular time point.
• 01:14:05As I move through time and
• 01:14:07export this data out,
• 01:14:08I'll get the data you know on a per
• 01:14:10object per track basis and be able
• 01:14:12to kind of export that into an Excel
• 01:14:14document and get get the tracking of
• 01:14:16these objects over time within the data set.
• 01:14:19Now a little bit better data set
• 01:14:21to kind of show that real quick.
• 01:14:22So demo another demo data set here.
• 01:14:24This is kind of really not a very
• 01:14:27great data set, but it's just a,
• 01:14:29it's an R18 demo here.
• 01:14:31The resolution in this one
• 01:14:33is really terrible.
• 01:14:33It's a really low resolution XY data set,
• 01:14:36but probably acquired really quickly.
• 01:14:39But the idea is that we can come up here,
• 01:14:41we don't need to classify
• 01:14:43this particular object.
• 01:14:44Spot detection,
• 01:14:44you can see it does a pretty good
• 01:14:46job of detecting all the spots.
• 01:14:47You can adjust thresholds,
• 01:14:48whatever it needs to be to kind of
• 01:14:51detect all the spots, come here,
• 01:14:54do the algorithm and you get
• 01:14:56your tracks over time.
• 01:14:57So now you can kind of see
• 01:14:59these guys tracking over time.
• 01:15:00If you click on one of these
• 01:15:03guys here and we focus it on it,
• 01:15:06right,
• 01:15:06he can see that guy tracking
• 01:15:09and moving through time,
• 01:15:12right? And then when you look at
• 01:15:14the statistics, you have all those
• 01:15:16statistics that are available for
• 01:15:18tracking that particular guy,
• 01:15:19whether it's looking at speed or anything,
• 01:15:21you can see the change in speed over
• 01:15:23time of that particular object. Visually,
• 01:15:25we can make this a little bit more pretty.
• 01:15:28You can come up here,
• 01:15:29You can do what's called the Dragon Tail.
• 01:15:30So I'm going to show a couple
• 01:15:31of time points at the end.
• 01:15:33So you kind of just see the the tail end
• 01:15:34of it as it's going for longer time lapses.
• 01:15:36This makes it really nice to
• 01:15:38kind of visualize the data sets.
• 01:15:39There's a lot of other visualization
• 01:15:41tools with tracks.
• 01:15:42Again, if you're doing any tracking
• 01:15:43data sets and you have any questions,
• 01:15:45set up a time we can talk about the
• 01:15:47different ways of of quantifying the data,
• 01:15:49visualizing the data,
• 01:15:51plotting the data, that sort of thing.
• 01:15:53We don't have a whole lot of
• 01:15:54plotting tools in the Mars.
• 01:15:54There's a Vantage tool here.
• 01:15:56Vantage tool is a nice little
• 01:16:00tool to render objects and make.
• 01:16:02I like to,
• 01:16:03I like to call it like kind
• 01:16:06of a data navigation tool,
• 01:16:09looking at the statistics,
• 01:16:11plotting your statistics a lot of times.
• 01:16:13For example,
• 01:16:13let's just say we look at the CD 80 and
• 01:16:15there's a lot of different objects in here.
• 01:16:17Say we look at the CD 86 that we've
• 01:16:20segmented out and you're like,
• 01:16:21wow, that's a lot of lot of
• 01:16:22stuff going on in there.
• 01:16:23But maybe if I open up Vantage now,
• 01:16:25Vantage is going to be a little bit messy
• 01:16:27because I have a lot of objects in here.
• 01:16:28So let me just one of the drawbacks
• 01:16:30here that turns everything on.
• 01:16:34But if we come in here and I
• 01:16:38just turn on CD86 for example,
• 01:16:42right, this is a 1D plot for CD86,
• 01:16:44I can say, OK, you know,
• 01:16:46what is the volume like I can look at
• 01:16:49here and it's like there's the right,
• 01:16:51there's the distribution of the volume of
• 01:16:53all those little puncture that I created
• 01:16:55looking at the the volume distribution.
• 01:16:57So this is just a 1D plot
• 01:16:58of that volume distribution.
• 01:16:59You can choose any statistic that you want.
• 01:17:01You can even filter the ones that
• 01:17:03are small or some filter the ones
• 01:17:04that are bright if you want.
• 01:17:06So you come up here to mean intensity
• 01:17:09channel #1 and you could say,
• 01:17:10hey you know what,
• 01:17:10these are the only ones that are bright or
• 01:17:11I only want to do the ones that are dim.
• 01:17:13These are the guys that are dim and you
• 01:17:15can you can plot little graphs of these
• 01:17:18guys as well within kind of a 1D plot.
• 01:17:20You also have a 2D plot.
• 01:17:222D plot gives you a little bit
• 01:17:25of flexibility here in terms
• 01:17:28of being able to plot,
• 01:17:30for example,
• 01:17:32volume versus
• 01:17:35sphericity. I don't know, right.
• 01:17:38It's just a way of looking at your data.
• 01:17:40These are those puncta that we're looking
• 01:17:42at and you can kind of distribute,
• 01:17:44you know, do any kind of graph
• 01:17:45within within the structure.
• 01:17:46The data that you're exporting
• 01:17:47is all down here.
• 01:17:48So if you did this export,
• 01:17:50you have your volume and your sphericity.
• 01:17:52If you hit this little save button on here,
• 01:17:54you'll export these guys side by side.
• 01:17:55Then you'll get the volume and the
• 01:17:57sphericity of every object that's plotted
• 01:17:59and detected within your surface scene.
• 01:18:01And again, it's just a way of navigating
• 01:18:03the data to kind of see is there a trend,
• 01:18:06trend where as the volume gets bigger,
• 01:18:07it gets more spherical or maybe it gets more,
• 01:18:11maybe the ones that are spherical
• 01:18:13or more overlapped, right.
• 01:18:14So you can look at the overlap value
• 01:18:15here and say, OK, well, there's,
• 01:18:17you know, volume versus overlap.
• 01:18:19So if it's a bigger punctum,
• 01:18:20maybe they're more overlapping,
• 01:18:21maybe they're less overlapping,
• 01:18:22maybe there's different morphology
• 01:18:24values there that we can we can
• 01:18:26look at or maybe it's look at
• 01:18:27intensity or something like that.
• 01:18:28There's a lot of different ways of
• 01:18:30looking at these data and kind of
• 01:18:32plot in these ways and structures.
• 01:18:34And then if you wanted to get really fancy,
• 01:18:35there is a 3D view of the same structure
• 01:18:38where you can plot in 3D these structures.
• 01:18:42So here you select the CD which
• 01:18:46is the surface object here,
• 01:18:51right? So again,
• 01:18:51we have all these guys turned off,
• 01:18:53so we just turn these guys off. Here.
• 01:18:56Again, I hate how it turns everything on.
• 01:18:58Makes it hard to only do one,
• 01:19:01but we just make sure.
• 01:19:02I usually don't have this many
• 01:19:04surfaces within my structure anyway.
• 01:19:06But anyway, you have this guy turned on.
• 01:19:08You go to the next step here and you can plot
• 01:19:10three statistics if you want four statistics,
• 01:19:135 statistics.
• 01:19:14You know color coding, right?
• 01:19:16So I can do XYZ and then I can do,
• 01:19:19you know, volume.
• 01:19:20So now it's color-coded by volume,
• 01:19:22but I I can do XY or I can do surface area,
• 01:19:27whatever, and it gets really complicated.
• 01:19:29Again, I think the more complicated you get,
• 01:19:31the harder it is.
• 01:19:32But you can plot whatever
• 01:19:33statistics you want here to kind
• 01:19:35of get an idea of what's going on.
• 01:19:36Viewing it in 3D and you
• 01:19:38get these structures here,
• 01:19:39So these are surfaces.
• 01:19:40So it's actually nice thing about Vantage is
• 01:19:42it's actually plotting the actual surfaces.
• 01:19:44So you don't you're not seeing a dot,
• 01:19:46you're seeing the actual
• 01:19:48surface object in 3D here.
• 01:19:50These are the actual surface
• 01:19:51objects relative to each other.
• 01:19:52So the small guys are over here,
• 01:19:53the big guys are over here.
• 01:19:55So you can actually see them side
• 01:19:57by side rendered in the structure.
• 01:19:59So like if you have guys and you're
• 01:20:02doing a plot here and looking at
• 01:20:04you know different parameters here,
• 01:20:06you can look at these colors,
• 01:20:08you can look at the size of the structures,
• 01:20:09Oh yeah,
• 01:20:10look at all the small ones that
• 01:20:11are at this part of the structure.
• 01:20:12It's nice, kind of.
• 01:20:13Could be a visual tools kind
• 01:20:15of show certain things.
• 01:20:17You know,
• 01:20:17to a lab meeting or your
• 01:20:18colleagues or something to kind
• 01:20:19of show these processes there.
• 01:20:21So Advantage can be a
• 01:20:22really powerful little tool.
• 01:20:22There also is a gallery tool here,
• 01:20:24plus every single punked up by itself.
• 01:20:27Kind of cool.
• 01:20:29Again, you can color code it as well.
• 01:20:31Put in here and say, oh,
• 01:20:32let's color code it based on sphericity.
• 01:20:36There you go, right?
• 01:20:38Not only is it by volume or area,
• 01:20:40but now it's color-coded
• 01:20:41which ones are spherical.
• 01:20:42So the ones that are purple
• 01:20:44are the least spherical,
• 01:20:45The ones that are red are the most spherical,
• 01:20:49right.
• 01:20:49You can see that here and
• 01:20:50you can plot that that's.
• 01:20:51So that's kind of an interesting way
• 01:20:53to kind of visualize every single bunk
• 01:20:54that kind of in a kind of ordered array.
• 01:20:56It's a nice little feature of the
• 01:20:58software to kind of do that if you
• 01:21:00wanted to do something like that.
• 01:21:03Finally, go back to Surf Beth mode,
• 01:21:06just to kind of wrap it up here,
• 01:21:10the idea of the snapshots I mentioned,
• 01:21:13right, Whatever you see on the screen,
• 01:21:15that's what you get.
• 01:21:16So if I have a screen here,
• 01:21:17whether I have a surface on or not,
• 01:21:20right, that Surface is going to be a snapshot
• 01:21:23of this image will return, you're right.
• 01:21:26So I do something like that.
• 01:21:27That's going to be part of the image.
• 01:21:29Again, my scale bar.
• 01:21:30Wherever my scale bar is,
• 01:21:31that's going to be on my scale bar.
• 01:21:32You can change the background.
• 01:21:34Right now my background is black,
• 01:21:37mainly because I have the volume turned on.
• 01:21:39If I turn the volume off,
• 01:21:40you'll see I have this kind of
• 01:21:42gradient background from brown,
• 01:21:44from Gray to white.
• 01:21:46Again, if you want to change any of the
• 01:21:47display properties under preferences,
• 01:21:49you're going to go over here to
• 01:21:51display and you can turn on the
• 01:21:53background to whatever color you want.
• 01:21:55I use this Linear Progress for blending.
• 01:21:57It's my favorite.
• 01:21:58I think it looks pretty cool.
• 01:21:59Gives you a little bit of depth if you will,
• 01:22:02using this kind of blending.
• 01:22:03Some people don't like it.
• 01:22:04You can turn it off to a single color.
• 01:22:05Gray.
• 01:22:06I do it like a light light
• 01:22:08dark Gray is my favorite.
• 01:22:10Black is a little bit too black,
• 01:22:11white is way too white.
• 01:22:13This is a nice little gradient
• 01:22:15in between view of of the data
• 01:22:19in terms of visualization, right?
• 01:22:20Like I said, it is what you
• 01:22:21get if I'm rotated like that.
• 01:22:23That's going to be my snapshot
• 01:22:24wherever my scale bar is.
• 01:22:25It's going to take a picture of
• 01:22:27this whole image view that we
• 01:22:29have here In terms of animation,
• 01:22:33animation is pretty straightforward.
• 01:22:34It's a it's a keyframe animation.
• 01:22:36Again, same deal.
• 01:22:37Whatever you see on the screen
• 01:22:39is what you get. So for example,
• 01:22:40if I wanted to just take this image
• 01:22:43right here as it is and do a 360 horizontal,
• 01:22:45I can click this button,
• 01:22:47it's going to add these little keyframes.
• 01:22:48You can see there's a A-frame here,
• 01:22:50A-frame here, A-frame here, A-frame here.
• 01:22:52And if I click on here and I rotate,
• 01:22:54you'll see it's going to just
• 01:22:56rotate it horizontally 360° because
• 01:22:58I clicked that 360° horizontal
• 01:23:00automatically boot that in there.
• 01:23:02If I click here,
• 01:23:03I can change the frames from 500 to
• 01:23:051000 means it's going to be much slower.
• 01:23:07The movement and turning is
• 01:23:08going to be slower.
• 01:23:09It's going to be a little bit smoother.
• 01:23:11It's going to take a
• 01:23:12little bit longer to play,
• 01:23:13could be a longer movie now depends on how
• 01:23:16many frames per second you play the movie.
• 01:23:18That's when you click this button here,
• 01:23:20it'll give you an option here to say how many
• 01:23:22frames per second do we want it to play.
• 01:23:24So 24 is pretty much our default.
• 01:23:26You can set it to whatever you want.
• 01:23:29And then when you play the movie,
• 01:23:30you can kind of predict how long that
• 01:23:31movie is going to last based on how
• 01:23:33many frames you have and the frames
• 01:23:34per second that it's going to play.
• 01:23:35So if you wanted to play for
• 01:23:37a specified amount of time,
• 01:23:38you can figure that out from your
• 01:23:41data set if you want to come in here
• 01:23:43and modify into these structures.
• 01:23:44Again, I can come up over here and say,
• 01:23:46OK, what I want to do 360 horizontal,
• 01:23:48but at this frame here,
• 01:23:49I want to zoom in to this guy,
• 01:23:51maybe be rotated a little bit
• 01:23:53that I can modify.
• 01:23:55I'm just modifying that keyframe so it's
• 01:23:57going to still do everything the same.
• 01:23:59I hit play here,
• 01:24:00it's going to go through the data
• 01:24:02set and you can see as it's turning,
• 01:24:03it's going to actually,
• 01:24:05you know,
• 01:24:05interpolate the zoom in,
• 01:24:06zoom into my data set now and it's
• 01:24:09going to zoom back out as it continues
• 01:24:11the rotation and keep turning
• 01:24:13around the data set to the end.
• 01:24:15So that's a really cool way to kind of,
• 01:24:17you know, make a rotation,
• 01:24:18maybe zoom in on a particular
• 01:24:20cell and zoom out.
• 01:24:21I typically would recommend people to
• 01:24:23kind of do the movements very simply.
• 01:24:26Don't make it complicated.
• 01:24:28The more complicated you get,
• 01:24:29the more chances are you're going
• 01:24:31to kind of have a messy movie that's
• 01:24:32going to give you a headache that's
• 01:24:34going to be hard to follow and hard to edit.
• 01:24:36Keep the movie relatively simple
• 01:24:37within the data set.
• 01:24:39You can add these frames manually as well.
• 01:24:42So delete all these guys and set
• 01:24:44it like this,
• 01:24:45and I can set this as my primary and it
• 01:24:48sets the primary and the end as the same.
• 01:24:50So it kind of loops on itself.
• 01:24:51But let's just say I just want
• 01:24:53to zoom into this guy here and do
• 01:24:54something like that, add a keyframe,
• 01:24:56maybe that's all you want to do.
• 01:24:57So it'll it'll just come.
• 01:24:58It'll just come from here.
• 01:25:00Stop it. Right.
• 01:25:01So it'll it'll start at this point,
• 01:25:02it'll zoom in and then it'll zoom out.
• 01:25:05Pretty, pretty straightforward.
• 01:25:06If I wanted to add another keyframe in
• 01:25:09the middle here and say you know what,
• 01:25:11I wanted to turn on,
• 01:25:13I wanted to turn on there those turn that on,
• 01:25:18hit add keyframe.
• 01:25:19So now as I go into here, it'll zoom in,
• 01:25:21when it gets to that keyframe,
• 01:25:23those guys will turn on and then
• 01:25:25when I get back to this keyframe,
• 01:25:26they'll turn off, right?
• 01:25:27And so you can modify this however you
• 01:25:29see what you can move these guys around.
• 01:25:31So if you wanted to go really quick,
• 01:25:33you can do this, right?
• 01:25:36And so if I play this movie here,
• 01:25:39it'll, you know,
• 01:25:42it'll go really quick between the two,
• 01:25:45right to there and then we'll kind
• 01:25:46of go nice and slow and move there.
• 01:25:48So it's just based on how many frames
• 01:25:50are in that particular data set to
• 01:25:52kind of control the speed and the size.
• 01:25:55We typically save as an MP4 file.
• 01:25:56That's our default using this
• 01:26:00codec that we have here.
• 01:26:01It gives us the best stability
• 01:26:02across different platforms, Mac, PC,
• 01:26:04PowerPoint, those sort of things.
• 01:26:06Does a pretty good job of being
• 01:26:08compatible with both applications
• 01:26:09where you're going to play the movie
• 01:26:11and and and be in the movie again.
• 01:26:13What you see is what you get.
• 01:26:15Sometimes people are like,
• 01:26:16well, I don't want that frame,
• 01:26:17I don't want these numbers.
• 01:26:19I I like the frame,
• 01:26:19but I don't want the numbers
• 01:26:21go into the frame and modify.
• 01:26:23Because even if I come up here and
• 01:26:24and I turn off the box and I turn off
• 01:26:26the grid and I turn off the ticket
• 01:26:27Marks and I turn off the access labels,
• 01:26:30my keyframes still has them all in there
• 01:26:32and it should oh actually no interesting,
• 01:26:36didn't save it.
• 01:26:37OK,
• 01:26:37so make sure you turn those off when
• 01:26:40you make your keyframes because they
• 01:26:42are going to be part of that keyframe.
• 01:26:44And if you don't want them in there,
• 01:26:45turn them off and then set those keyframes
• 01:26:48so that you don't have those movies on again.
• 01:26:50Same with the the scale bar.
• 01:26:52The scale bar is going to move.
• 01:26:54Some people get a little
• 01:26:55bit freaked out by that,
• 01:26:57but as you're zooming in,
• 01:26:59the scale bar will adjust and you'll
• 01:27:00see it kind of tweak a little bit.
• 01:27:02Sometimes the scale bar as
• 01:27:03you're rotating it, see how it,
• 01:27:05I don't know if you noticed there,
• 01:27:06see how how I'm rotating it,
• 01:27:07that scale bar changes.
• 01:27:09That's normal.
• 01:27:10That's because the scale
• 01:27:12bar is based on the center
• 01:27:14of the volume. And so when you're
• 01:27:16visualizing the center of the volume,
• 01:27:19that scale bar is based on
• 01:27:21the center of the volume.
• 01:27:22We view our data in what's
• 01:27:23called perspective mode.
• 01:27:24Just so you kind of get this out there,
• 01:27:27we have two modes.
• 01:27:28There's perspective mode and orthogonal mode.
• 01:27:30Almost 90% of the time
• 01:27:32you're in perspective mode,
• 01:27:33especially with the 3D data set,
• 01:27:34mainly because you want to get a little
• 01:27:36bit of a better 3D feel of your file.
• 01:27:38Which means the files in the front
• 01:27:39are going to look bigger than the The
• 01:27:41cells in the front are going to look a
• 01:27:43little bit bigger than the cells in the back,
• 01:27:44but they're the same size.
• 01:27:46But again, it's just that perspective
• 01:27:48showing that they're, they're different.
• 01:27:49But the scale bar is going to be based on,
• 01:27:52oops, the scale bar is going to be
• 01:27:54based on the center of the volume,
• 01:27:55based on the camera.
• 01:27:56So it's going to be based
• 01:27:58on somewhere in there.
• 01:27:58That's the that's the size,
• 01:27:59Wherever your center is,
• 01:28:00that's the size of the structure.
• 01:28:02So as you rotate this guy,
• 01:28:03you'll see it moving a little bit.
• 01:28:04Depending on what zoom library you're in,
• 01:28:06you might see that being modified.
• 01:28:08That is normal.
• 01:28:09That's often usually a telltale
• 01:28:10sign that it's in a Mars movie.
• 01:28:12Sometimes people post things on Facebook
• 01:28:14and YouTube and things like that.
• 01:28:16I'm like, Oh yeah,
• 01:28:17that's in a Mars movie.
• 01:28:17I can tell because of the scale bar.
• 01:28:20That's just something we've done people,
• 01:28:22because we have to put a
• 01:28:23scale wall somewhere with a 3D
• 01:28:25perspective mode data set.
• 01:28:25So that's just what we've been doing.
• 01:28:28And again,
• 01:28:28it's the best estimate that we can can
• 01:28:31give in terms of the the 3D data set.
• 01:28:35OK, well, I think I'm going to stop there.
• 01:28:39I think I don't know if Matthias,
• 01:28:41if you're still there,
• 01:28:43if you want to say anything.
• 01:28:44I don't know if I see any
• 01:28:47questions specifically. Yeah.
• 01:28:50So, yeah, there were a bunch of
• 01:28:53questions on the that people put in
• 01:28:56as you went through the presentation.
• 01:28:58So you cannot see them.
• 01:29:00I can read them. I am looking for them.
• 01:29:04Oh, wait, hold on. It's minimized.
• 01:29:06How do I? Yeah, let me try that.
• 01:29:10Oh, yes, there they are.
• 01:29:11Excellent. OK, so I would start,
• 01:29:14I mean, the first question
• 01:29:15that came in was from Caroline.
• 01:29:16So she's asking,
• 01:29:17when you manually place spots in a 3D image,
• 01:29:20is there a way to make them
• 01:29:22the same size automatically?
• 01:29:24Whenever I place a spot and change its size,
• 01:29:26the next spot I place would not
• 01:29:28have the same size and I need
• 01:29:30to manually resize every spot.
• 01:29:31Yeah.
• 01:29:33So typically what I do when
• 01:29:37you're manually placing spots.
• 01:29:38So let's look at this example here.
• 01:29:40That's a great question.
• 01:29:42Let me
• 01:29:45turn off this here. Right. So
• 01:29:54where is that? Where is it Gray?
• 01:30:05Oh, I changed the color. OK,
• 01:30:09yeah, so so let's just say I detected
• 01:30:11all these spots manually and I wanted
• 01:30:14to add another spot of the same size.
• 01:30:16Say I wanted to add a spot right here,
• 01:30:17but I wanted the same size of this
• 01:30:19spot so often. Sometimes, yeah,
• 01:30:21if you don't change the size,
• 01:30:23usually it's going to be the same size.
• 01:30:24But you can come up here and say, OK,
• 01:30:26well I can come up here and I can,
• 01:30:27I can add a spot, whatever size I want.
• 01:30:28I can come up there and add a spot,
• 01:30:29But that's not obviously not
• 01:30:31the same size spot.
• 01:30:32You can hold down Shift and you
• 01:30:33can get rid of those spots as well.
• 01:30:35Shift click is to add the spot.
• 01:30:37But my rule of thumb is if you have some
• 01:30:39spots that are already been created,
• 01:30:41if you click on one of the spots,
• 01:30:44the size of the box automatically
• 01:30:46sets to the size that that spot is.
• 01:30:48Then if I come up here and add that spot,
• 01:30:50that should be the the exact same spot size.
• 01:30:54That's the the easiest way to do it.
• 01:30:56Changing the spot size manually is a
• 01:31:00pain because we can't do it as a group,
• 01:31:02it has to be done one at a time.
• 01:31:04But that that trick works pretty well.
• 01:31:06I I think that's probably the easiest
• 01:31:08way to kind of get that solution.
• 01:31:10Hopefully that addresses your question.
• 01:31:13So do you want me to read it?
• 01:31:15So how do you want to do?
• 01:31:17I can read it. OK, so yeah, I would go.
• 01:31:19I just got to expand it and get it out here.
• 01:31:22Does converting into Mars file
• 01:31:23change the original data set?
• 01:31:25No, it does not. So our conversion is lot.
• 01:31:29It's a lossless file conversion.
• 01:31:31So the the raw data, the pixel intensities,
• 01:31:34the voxel sizes,
• 01:31:35everything is exactly the same,
• 01:31:37The intensities are all the same.
• 01:31:39So if you if you acquire it as
• 01:31:41a 16 bit data set,
• 01:31:42the Mars file will be a 16 bit
• 01:31:43data set if it's a 32 bit.
• 01:31:45If it's a 8 bit, Mrs.
• 01:31:46will convert that file and be
• 01:31:48exactly the same file.
• 01:31:50Our file format is a multi
• 01:31:52resolution HDF 5 format.
• 01:31:54I'm sure that doesn't mean a
• 01:31:55whole lot to a lot of you,
• 01:31:56but it is a it is a method of visualizing
• 01:31:59the data with multi resolution levels.
• 01:32:01So what that means is when you're
• 01:32:03visualizing a really big data set,
• 01:32:04I don't know if I have one here,
• 01:32:06but if you're visualizing a really
• 01:32:07big data set, as you're zooming in,
• 01:32:10you're only, you know,
• 01:32:11basically you can kind of see
• 01:32:12it in this image a little bit.
• 01:32:14If I turn on the image here and I look
• 01:32:18at it at kind of a low resolution,
• 01:32:20you can kind of see the data,
• 01:32:22but in here it looks pretty dark.
• 01:32:24That's because the pixels that
• 01:32:25you're looking at there,
• 01:32:26you can't see them yet.
• 01:32:27They're so small based on the
• 01:32:29pixels on your screen, right,
• 01:32:30your pixels on your monitor,
• 01:32:31whether it's a high resolution
• 01:32:33monitor or low resolution monitor.
• 01:32:34As you zoom in,
• 01:32:35you're not going to see those pixels
• 01:32:37until you get to a particular zoom
• 01:32:38level where the size of your pixel
• 01:32:40on the screen at least matches the
• 01:32:42size the pixel on your monitor, right.
• 01:32:44And so as I zoom in,
• 01:32:45you're going to see,
• 01:32:46I'm just going to do it stepwise.
• 01:32:47Some of those the noise kind of
• 01:32:49starts seeing it higher and higher
• 01:32:50as you do more and more and it's
• 01:32:52at some point you're not going to,
• 01:32:54it's not going to change much because
• 01:32:55you're at the highest resolution level,
• 01:32:56but we're visualizing just the
• 01:32:58resolution level that you can
• 01:33:00see at the zoom and the cameras
• 01:33:02room that you have within.
• 01:33:03But yes,
• 01:33:04it's a lot totally lossless creation.
• 01:33:06So don't you don't have to worry
• 01:33:08about any of any any loss of signal
• 01:33:14problem of video. Yeah.
• 01:33:16These these I think was earlier.
• 01:33:19Oh, with the OH with the video.
• 01:33:21OK All right. So, Caroline,
• 01:33:23you have another question here.
• 01:33:25What is the best way to import EM data?
• 01:33:27Do I need to invert it first so it's
• 01:33:28all black, contrast becomes white.
• 01:33:30So Mars sees it as fluorescence.
• 01:33:32So EM data is interesting.
• 01:33:35So EM data typically is not fluorescent
• 01:33:39depending on how you've acquired it.
• 01:33:41I've done EM way back in the
• 01:33:43past and converted our our images
• 01:33:46to a digital digital series.
• 01:33:48I did some EM serial sectioning
• 01:33:50back in the day.
• 01:33:51From what I remember,
• 01:33:53they were monochromatic images
• 01:33:55from the camera.
• 01:33:56They were not,
• 01:33:58they were not
• 01:34:01RGB images. So we just monochromatic.
• 01:34:06I would just keep it as a
• 01:34:09monochromatic TIFF probably.
• 01:34:10Now Mars does not do any alignment.
• 01:34:12So if you're looking like for serial
• 01:34:15sectioning and things like that,
• 01:34:18they they need to be automatically or already
• 01:34:20aligned before you bring them into Mars.
• 01:34:22We used to have a,
• 01:34:23we used to have an aligner way,
• 01:34:24way back in the day.
• 01:34:25They stopped developing it a while ago.
• 01:34:27They'd never kind of re
• 01:34:28implemented it as a tool.
• 01:34:29It might happen at some point in the future.
• 01:34:31Right now you'd have to go
• 01:34:33somewhere else to do the alignment.
• 01:34:34Fiji has a couple great tools.
• 01:34:36I'm sure there's other tools out there,
• 01:34:37maybe with how you acquired the data,
• 01:34:40but I think
• 01:34:43I mean if you want to invert it, you can.
• 01:34:47But usually it's a monochromatic
• 01:34:50image with black. You know,
• 01:34:52kind of a grayscale grayscale image.
• 01:34:54You can still use that grayscale image and
• 01:34:56use the machine learning and and train it.
• 01:34:58That's definitely doable.
• 01:34:59I've done it on a couple occasions
• 01:35:01where you can use that pixel
• 01:35:04classification to do segmentation for
• 01:35:05EM data or you can do some contour
• 01:35:09tracing to create surfaces on EM data.
• 01:35:11I didn't go into the contour tracing tool.
• 01:35:14If you are interested in something
• 01:35:15like that reach out to me.
• 01:35:16I have a video tutorial I can
• 01:35:17send you about how to do contour
• 01:35:19tracing inside of a Morris,
• 01:35:20but that's a very hands on slice
• 01:35:22by slice kind of tracing of your
• 01:35:24data and then it creates a 3D
• 01:35:26volume based on based on that.
• 01:35:31What does the smooth function actually
• 01:35:33do the data to make it look smoother?
• 01:35:35It's a Gaussian filter as I said when I was
• 01:35:38trying to present you earlier. When you do,
• 01:35:43when you do the smoothing step here,
• 01:35:47if you look in. Also just a rule of thumb,
• 01:35:49if you ever have a question on anything
• 01:35:51in a Morris and you don't want to
• 01:35:53call support and your tech people or
• 01:35:54people in your lab are not around,
• 01:35:56that can't help you and you
• 01:35:58can't reach me right away.
• 01:35:59Usually calling me is not
• 01:36:00the best way to contact me.
• 01:36:02Usually by e-mail is the
• 01:36:02best way to contact me.
• 01:36:03So I'm not,
• 01:36:04I'm not someone that's going to be
• 01:36:05able to answer the phone when you're
• 01:36:07sitting in front of the microscope,
• 01:36:08very often much in front
• 01:36:09of the imaging computer.
• 01:36:10So a lot of times you want to try to find
• 01:36:12the answer yourself and send me an e-mail.
• 01:36:14But a lot of times in Amaris,
• 01:36:15all the Wizards in Amaris,
• 01:36:17hopefully everywhere.
• 01:36:18Most of us, if you find an error,
• 01:36:19let me know.
• 01:36:20But if you right click on anywhere in Amaris,
• 01:36:22like for example,
• 01:36:23if I right click right here
• 01:36:24on Smooth and I hit show help,
• 01:36:26it's going to open up our HTML
• 01:36:31manual and it'll show you.
• 01:36:32So right here it'll tell you exactly
• 01:36:34what the smoothing factor does.
• 01:36:36It's basically a Gaussian filter.
• 01:36:38That's all it is but anywhere.
• 01:36:40Tomorrow if you can click on
• 01:36:41here to get what that is and it
• 01:36:42even tells you exactly what the
• 01:36:43background subtraction does.
• 01:36:44So you can have a if you want to
• 01:36:46try to write up a methods back
• 01:36:47paper or kind of tell somebody
• 01:36:48what you how you process your data.
• 01:36:50It tells you exactly what what these
• 01:36:52processes do within within the software.
• 01:36:56OK, so can this training be
• 01:37:01applied to multiple images?
• 01:37:03I assume you're talking about
• 01:37:05the pixel classification.
• 01:37:06So that's a great question.
• 01:37:08I didn't really get into that.
• 01:37:09And again,
• 01:37:09I'm going to,
• 01:37:10there are some video tutorials
• 01:37:12available about kind of best
• 01:37:14practices for doing pixel
• 01:37:15classification and we are building
• 01:37:17our library of tools out there.
• 01:37:19There's a bunch of tools out there.
• 01:37:21Just to show you real quick here,
• 01:37:23there is a tool here called the Mars Bites.
• 01:37:26It's kind of developed by
• 01:37:31Mars Bites. This is a YouTube page,
• 01:37:33so if you search Mars Bites on YouTube,
• 01:37:35there are right now 30 videos,
• 01:37:38short little videos,
• 01:37:38all about 5 to 6 minutes,
• 01:37:407 minutes long, some shorter,
• 01:37:42doing all different various
• 01:37:44small little features of a Mars.
• 01:37:47A lot of the stuff I didn't even cover
• 01:37:48today because they're they're not really.
• 01:37:50They're more kind of offhand type things,
• 01:37:53you know, adding a surface
• 01:37:54with the magic wand, Z series,
• 01:37:56animation, labels and vantage.
• 01:37:57There's a lot of different things
• 01:37:58that are here that you.
• 01:37:59Oh yeah, look, that might be interesting.
• 01:38:00I'm going to watch that short little videos,
• 01:38:02nice little resource for you to go in here.
• 01:38:06I'm sorry, what was the last
• 01:38:07question on the question?
• 01:38:08Oh, for multiple trainings.
• 01:38:10So when you train your data sets,
• 01:38:13typically the pixel classification,
• 01:38:15if I come up on here and I say,
• 01:38:17hey, I'm going to,
• 01:38:21I'm going to just do it,
• 01:38:22I'm going to do it on one of these data
• 01:38:23sets because these are all the same size,
• 01:38:25I think. No, no, they're not.
• 01:38:28So the idea here is like this is a,
• 01:38:30this is a control data set, right?
• 01:38:31This is a three, three data sets.
• 01:38:33Let's just say I want to do my
• 01:38:35pixel classification training
• 01:38:36on one of these files, right.
• 01:38:38I might open this file up here and say,
• 01:38:42hey, I want to,
• 01:38:43I want to train my pixel classification.
• 01:38:44I'm going to come in here,
• 01:38:46I'm going to go and make
• 01:38:48these machine learning.
• 01:38:49I'm going to train the green channel.
• 01:38:51Come in, here I go.
• 01:38:53You know I'm not going to do this
• 01:38:54very good here to background and I do
• 01:38:59foreground or whatever.
• 01:38:59And you train and predict, right?
• 01:39:01And you get a result.
• 01:39:01You're going to get some sort of some
• 01:39:04sort of training of that data set.
• 01:39:06I can save that training and I can apply
• 01:39:08that exact training to multiple files.
• 01:39:11I can. Like I said,
• 01:39:12if I if I finish this training and say,
• 01:39:13hey, I want to save this training here
• 01:39:15and I do store parameters for batch,
• 01:39:18I can call this green with pixel
• 01:39:21classification whatever, right?
• 01:39:22And so now I can go. I can,
• 01:39:25I can go here and open up another data set.
• 01:39:30Right. Come up here.
• 01:39:32There's a favorite creation
• 01:39:34parameters here I can do,
• 01:39:36oh, what did I call it?
• 01:39:42Green with fixed classification.
• 01:39:43Right. Hit next.
• 01:39:44It's going to use the same exact training.
• 01:39:47The data is exactly the same.
• 01:39:49It has to be from this, you know,
• 01:39:50Everything has to be exactly the same.
• 01:39:51But now you have your surface creation
• 01:39:53is the same exact classification in
• 01:39:55that particular data set, right?
• 01:39:56Same exact.
• 01:39:57You can do this as many times as you want.
• 01:39:59You can even do this in a batch processing.
• 01:40:01If you have the batch module,
• 01:40:02you can go into the batch module.
• 01:40:03Now I didn't go into the batch module again.
• 01:40:06If you're interested in kind
• 01:40:07of being able to kind of do,
• 01:40:09you can set these parameters on a on a,
• 01:40:11set it and go kind of scenario where
• 01:40:13they're all very similar and the same
• 01:40:15exact parameters work well, then batch.
• 01:40:17Might be an option for you.
• 01:40:20I'm I'm a fan of kind of doing it
• 01:40:21one at a time and and making sure
• 01:40:23that it works well on these data
• 01:40:24sets before I go on to the next one.
• 01:40:26But again,
• 01:40:27batch processing here is kind of reusing
• 01:40:29the same parameters on this image and
• 01:40:31then you can kind of validate that result.
• 01:40:33Now one of the things that people will say is
• 01:40:36that I don't want to train on just one image,
• 01:40:38I want to do the training on multiple images.
• 01:40:40You can do that right?
• 01:40:42So if I trained it on that image
• 01:40:43and I open up this one, right.
• 01:40:45And I come up here and I was like,
• 01:40:46you know what?
• 01:40:47I'm going to use that green training,
• 01:40:50green cells with PC, right?
• 01:40:52I come up here and I do this.
• 01:40:54You see where it says keep training data.
• 01:40:56It's going to keep the training data.
• 01:40:58But I can come up here and I can
• 01:40:59look at this image and say, oh,
• 01:41:00you, you can look at this and say,
• 01:41:01oh, wow, that's terrible.
• 01:41:02I want to come up here.
• 01:41:03This is, I need to train this better.
• 01:41:05On this particular image,
• 01:41:06it didn't work real well.
• 01:41:07I can come up here and train some of
• 01:41:08these structures here and and retrain it.
• 01:41:10That's going to build on
• 01:41:11that existing training.
• 01:41:12So it's going to take the training.
• 01:41:13I did it on image #1,
• 01:41:15it's going to modify it on image #2,
• 01:41:17and so now I have a new
• 01:41:19surface creation on image #2.
• 01:41:20I finished this process right?
• 01:41:22And then I can come in here
• 01:41:25and say pixel classification.
• 01:41:27I can save and I'm going to
• 01:41:29say green with PC I just,
• 01:41:34I just like saving for two.
• 01:41:35Sometimes I say now that's
• 01:41:37second second round, right?
• 01:41:38So now I've trained it on one,
• 01:41:40I applied it to a second one.
• 01:41:42Doesn't quite work as well,
• 01:41:43but I'm going to train it on that one.
• 01:41:45Now I have this,
• 01:41:47this second,
• 01:41:47this kind of merged creation parameter
• 01:41:50that's based on two separate images,
• 01:41:52but it's done separately and I'm not a huge,
• 01:41:54big fan of that. Typically,
• 01:41:56what I would like to do to do proper
• 01:41:59training data set is to kind of merge
• 01:42:01these files together into one file.
• 01:42:03Now there's a couple different ways to
• 01:42:04do that. One is to do it in a time to
• 01:42:07make yourself a fake time lapse and
• 01:42:09then train this data set one at a time.
• 01:42:11Now I was hoping that you'd kind of build
• 01:42:12this into the software a little bit better,
• 01:42:14but they haven't quite done that yet
• 01:42:15because it's brand, brand spanking new.
• 01:42:17And so the idea is you want to be
• 01:42:20able to have a time lapse data set.
• 01:42:22So, but to do that they have
• 01:42:24to be exactly the same size.
• 01:42:25Unfortunately, these guys here,
• 01:42:28358 by three, 49356 by three,
• 01:42:31they're not exactly the same sizes,
• 01:42:32so we can't add them through time
• 01:42:34because they're different sizes.
• 01:42:35That's not always the case.
• 01:42:36Sometimes two data are exactly the same size,
• 01:42:391024 by 1024, blah blah blah,
• 01:42:41and 10 and 20 slices,
• 01:42:43but they're also different Z slices as well.
• 01:42:45So what I typically would do in a
• 01:42:48case like this, and I want to say,
• 01:42:50you know what, I have 10 data sets,
• 01:42:51but I want to train it on three
• 01:42:53different data sets.
• 01:42:53I don't want to just train it on one
• 01:42:55because there's a little bit of variation
• 01:42:57between one image and another image.
• 01:42:58So what I end up doing here is I would
• 01:43:01come up here and I'm just going to
• 01:43:04open up just the image here, right?
• 01:43:05So I have image #1, right?
• 01:43:07If I come up here and I do
• 01:43:10edit and I add slices,
• 01:43:12the size of the image doesn't matter.
• 01:43:14Well,
• 01:43:14actually it does matter because
• 01:43:15it has to be the same size.
• 01:43:16Shoot,
• 01:43:19OK, that's not going to work.
• 01:43:20Let me find a different example.
• 01:43:23Well, I'll just use this.
• 01:43:24I'll use the same file as an example.
• 01:43:27It still has to be the same XY dimensions
• 01:43:29to be able to do it this way as well.
• 01:43:31But the idea is that you're
• 01:43:32going to come up here, do edit,
• 01:43:34and you're going to either add time points.
• 01:43:36So if I do add time points here,
• 01:43:38and I take this image here and I add it,
• 01:43:41I'll have one time point here
• 01:43:42and I'll have another time point.
• 01:43:44Now it's the same time point, right?
• 01:43:45But you'll have one time .2 time points.
• 01:43:47And so when you create
• 01:43:49your creation parameters,
• 01:43:50you'll hit the blue tab and
• 01:43:51you can excuse me,
• 01:43:52you're going to train data on time .1,
• 01:43:55background and foreground.
• 01:43:56You're going to get time .2.
• 01:43:57You're going to train data
• 01:43:58time background and foreground.
• 01:43:59And you're going to optimize it on both
• 01:44:01until you're done and you're happy
• 01:44:03with the segmentation that's happening
• 01:44:04on both of those images at the same time.
• 01:44:06So that's one option.
• 01:44:08If they're the same,
• 01:44:11exactly the same size XY and Z,
• 01:44:13you can add them in time lapse.
• 01:44:15However, that's not always the case because
• 01:44:18a lot of times the ZS are different,
• 01:44:22X&X&Y is the same,
• 01:44:23there's still 1024 by 1024,
• 01:44:24but the ZS are different.
• 01:44:25So what I would do in a case like that,
• 01:44:27and it's a little bit counterintuitive,
• 01:44:28but I'm going to do add slices.
• 01:44:30So what that basically does is it
• 01:44:32all added on top of each other.
• 01:44:34So you'll get an image here and
• 01:44:36if I go into slice,
• 01:44:37you'll see it right?
• 01:44:38So here's your first image,
• 01:44:40and then as you go down,
• 01:44:41that image will disappear and then
• 01:44:42the second image will be at the top.
• 01:44:43And so the idea is that the pixel
• 01:44:46classification doesn't take into
• 01:44:47account the Z depth of the pixel,
• 01:44:49it just takes into account the pixels
• 01:44:51that are just surrounding that pixel.
• 01:44:53And so a lot of times you can do
• 01:44:55this trick by adding it in Z so
• 01:44:57that I can train pixels up here,
• 01:44:59train pixels down here,
• 01:45:002 totally different images.
• 01:45:01And then when I save that creation parameter,
• 01:45:03it's based on two separate
• 01:45:05images within the data set.
• 01:45:06That's typically what I would end up doing.
• 01:45:08You can also crop your other data sets
• 01:45:10to fit so that they are the same size.
• 01:45:13Or if you know you're going to
• 01:45:14do it on a training data set,
• 01:45:15acquire data in a way that you can
• 01:45:17kind of import them in so that
• 01:45:18this is your training data set.
• 01:45:20Here's an image, here's an image.
• 01:45:21They're exactly the same size,
• 01:45:22and if you knew you're going to
• 01:45:23do this in the future,
• 01:45:24then you can kind of make sure
• 01:45:26that you take them,
• 01:45:27take these training data sets
• 01:45:29with the same size, same Z step,
• 01:45:31that sort of stuff.
• 01:45:31And that way when you go to
• 01:45:33import them together and make your
• 01:45:34time time lapse is the easiest,
• 01:45:35the best way to do it.
• 01:45:36So if you're taking a 10/24 or 10/24,
• 01:45:38just take them two or three images
• 01:45:41with exactly the same number of slices,
• 01:45:43add them together in that time lapse
• 01:45:45and then do the training on those data
• 01:45:47sets for all the remaining images
• 01:45:48that you have in your data set.
• 01:45:50So that's typically what I would
• 01:45:51do in a case like that.
• 01:45:53So hopefully that answers that question.
• 01:45:59Can you split volumes post
• 01:46:00processing after making a volume?
• 01:46:02If, for example, escalating C
• 01:46:03point isn't accurate enough,
• 01:46:07The quick answer there is yes we can.
• 01:46:10It is a very crude measure unfortunately,
• 01:46:14so let me demonstrate that
• 01:46:16really quickly here. I'll do it.
• 01:46:19Let's do it on this page.
• 01:46:21So the the tool for cutting
• 01:46:23is a it's called a cut tool.
• 01:46:26So if we look at Oops,
• 01:46:31oh wrong, oh wrong one, sorry.
• 01:46:38So the cutting is a very crude
• 01:46:39it's like taking a hatchet when you
• 01:46:41really need a scalpel, but it works
• 01:46:43depending on what you're doing here.
• 01:46:45So if I were to say, say I wanted to cut
• 01:46:49this little tip off that you know what,
• 01:46:50that's definitely not part of the cell.
• 01:46:53There is a pencil tool here
• 01:46:55and there's a cut surface,
• 01:46:57you know, scrolling over here.
• 01:46:58Hold down shift and left click and
• 01:47:01you'll see a little blue line.
• 01:47:03The blue line is always going to be vertical.
• 01:47:05As you place that blue line,
• 01:47:06it is always vertical.
• 01:47:08So you rotate your image appropriately
• 01:47:09and that line gets placed vertically.
• 01:47:11Can't change it any other way.
• 01:47:13But then you can click cut surface
• 01:47:15and then that surface there now is
• 01:47:18separate from this guy and you can take
• 01:47:21this guy and you can get rid of it.
• 01:47:23So it it it's a it's a good tool,
• 01:47:26it gets the job done,
• 01:47:27but it's not a very sensitive tool.
• 01:47:29And just be aware if you do use the cut tool,
• 01:47:32if I cut like this,
• 01:47:33you can see unfortunately it's
• 01:47:35not just cutting this one segment.
• 01:47:37The way they make their surfaces in a
• 01:47:39Morris and the way they changed it,
• 01:47:41it's going to cut every single
• 01:47:42part of that surface together.
• 01:47:44So if I did something like that,
• 01:47:47you'll see it's going to cut all those
• 01:47:48little pieces and I can visualize that here,
• 01:47:50just show this real quick.
• 01:47:51Here Object ID,
• 01:47:52you can see these these pieces
• 01:47:54are all cut now they're separate.
• 01:47:57You can go back and unify them later
• 01:47:59so that they're more connected.
• 01:48:01So like, oh, I cut this off,
• 01:48:02but this is the one I want to get
• 01:48:03rid of that one.
• 01:48:04But these guys here I still want to keep,
• 01:48:06if I hold down control and select that,
• 01:48:07go to the edit, there is a Unify.
• 01:48:09I can unify that back.
• 01:48:10It's not going to change the volume
• 01:48:12because we didn't change the volume
• 01:48:13when we cut it and just split it,
• 01:48:14You're going to get it back
• 01:48:15to the original size.
• 01:48:16But then I can,
• 01:48:17I can select this guy here and
• 01:48:18get rid of that corner if I want
• 01:48:19to just get rid of that guy.
• 01:48:20So a lot of different editing tools.
• 01:48:22Like I said, it's like a hatchet,
• 01:48:23but it it'll it'll work in a pinch.
• 01:48:25If it's a very complicated surface,
• 01:48:26it's not going to be very useful.
• 01:48:28But if it's a simple thing,
• 01:48:29you want to cut off a cell or split
• 01:48:30a cell that didn't get split,
• 01:48:32you can use this cut tool to
• 01:48:34typically get get the the result
• 01:48:36that you're looking for.
• 01:48:38The next question I think was about
• 01:48:42individual colors of the microglia.
• 01:48:44Short answer there is.
• 01:48:47I did object ID so every surface
• 01:48:52gets a random color so it gets an
• 01:48:54object ID so that's base color
• 01:48:55object ID so
• 01:48:56every object you'll get a separate color.
• 01:48:58That's that's all I did there for that.
• 01:49:00You can visualize these
• 01:49:02surfaces with statistics coding.
• 01:49:04You can do that as well.
• 01:49:05So for example, if I wanted to get
• 01:49:07it based show based on volume,
• 01:49:08you can get oops sorry,
• 01:49:11you can get it to look like
• 01:49:16spectrum. The ones that are
• 01:49:17red are my bigger guys,
• 01:49:19guys are purple or more smaller guys.
• 01:49:20Another way of visualizing the data,
• 01:49:22giving it, you know, giving some
• 01:49:24statistical context to your volume,
• 01:49:25sometimes that's a useful
• 01:49:26way of doing it as well.
• 01:49:27But object ID is the quick and
• 01:49:28dirty way to kind of show them.
• 01:49:29So typically if I am cutting my
• 01:49:32cells and using that cut tool,
• 01:49:33I'm usually in the object ID because
• 01:49:35that'll tell me exactly if I cut it,
• 01:49:37it'll show me the colors really clearly.
• 01:49:39It's like, Oh yeah, look,
• 01:49:39cut that little piece off.
• 01:49:41I can go find that or I can undo it
• 01:49:42and and and merge them together later.
• 01:49:48OK. I'm going to try to go through all
• 01:49:50the questions here as best I can here so
• 01:49:52that we don't get through the end here.
• 01:49:54So I know we're a little bit past our time,
• 01:49:56but hopefully it looks like most
• 01:49:57people kind of stuck around.
• 01:49:58So great. I appreciate that.
• 01:50:01So, hi, Matthew.
• 01:50:02I'm wondering how to use that microblia
• 01:50:04image to do stroll analysis after it is
• 01:50:07reconstructed using the filament tool.
• 01:50:08So I think I mentioned that real briefly.
• 01:50:11Again, Fill in analysis is a totally
• 01:50:13separate seminar that we can talk about.
• 01:50:16But any kind of filament that you've
• 01:50:19generated, any filament object that
• 01:50:21you've created under the Statistics tab,
• 01:50:24the stroll analysis is right here.
• 01:50:27And so it does it.
• 01:50:28You'll see there's going to be a
• 01:50:30lot of data here, unfortunately.
• 01:50:31Sometimes it's a little bit hard to imagine,
• 01:50:33but there's a filament ID,
• 01:50:34so every filament has its own ID.
• 01:50:37If you wanted to kind of make
• 01:50:38it a little bit simpler,
• 01:50:39maybe you don't have as many
• 01:50:40in the structure,
• 01:50:41but if you kind of take this guy out
• 01:50:43and just duplicate it out on its own,
• 01:50:46then the Shoal analysis will
• 01:50:47kind of be all by itself, right?
• 01:50:49It'll be a little bit easier
• 01:50:50to kind of export and look at.
• 01:50:51But if you export that to Excel,
• 01:50:53that that value is there and you can see
• 01:50:57the intersections here one through 10,
• 01:51:00and it's doing it on one Micron intervals.
• 01:51:03So if you wanted to do every 10,
• 01:51:05you just have to go out and
• 01:51:06pull out every 10,
• 01:51:07It's probably the easiest way to do it.
• 01:51:08There is an option here to
• 01:51:10set the Shoal resolution.
• 01:51:13By default it's set to one,
• 01:51:16but if I change this to 10,
• 01:51:17it's not going to change my statistics here
• 01:51:19because I already created my filament.
• 01:51:20If I set this to 10 and then do my filament,
• 01:51:23then it's only going to show every 10.
• 01:51:25So I I tend to just like the 1 Micron
• 01:51:28view here and then export it to Excel
• 01:51:30and then just take every 10th one.
• 01:51:32It's pretty easy to do that in
• 01:51:34Excel and other applications to
• 01:51:35kind of get the the resolution that
• 01:51:37you want for for Shoal analysis.
• 01:51:39OK,
• 01:51:41is it possible to batch export one
• 01:51:44specific data from multiple image,
• 01:51:50one specific statistic out?
• 01:51:58You can't. Well, kind of.
• 01:52:01I guess this is a little bit
• 01:52:04of a tricky scenario here.
• 01:52:06Let me show you real quick here,
• 01:52:07see if I can demonstrate it on an example.
• 01:52:09Let's see if I'm going to hopefully,
• 01:52:11hopefully I'm going to answer your question.
• 01:52:12I know we don't have any audio for you guys,
• 01:52:15but basically what it seems like you're
• 01:52:18asking where let's go here, go here.
• 01:52:20I'm going to just get rid of
• 01:52:21all these guys getting away.
• 01:52:30OK, so if these are three cells,
• 01:52:333 images that I processed,
• 01:52:36if I come up here in this guy,
• 01:52:39you can see how there's a there's
• 01:52:41a surface that's called KO.
• 01:52:43They say It's A Knockout, whatever.
• 01:52:44Then here's my red signal, right?
• 01:52:46These are two separate surfaces.
• 01:52:48There's there's this guy here,
• 01:52:51nice and big, and there's this KO guy.
• 01:52:53That's this guy, right?
• 01:52:55Two separate surface objects with KO and red.
• 01:52:57And I saved that file, right?
• 01:52:59If I come back over here and I
• 01:53:01look at these images here, here,
• 01:53:03there's a KO in this image, right?
• 01:53:06I've made that surface I don't have.
• 01:53:08I don't have the red in here,
• 01:53:09but I have the KO surface in here.
• 01:53:10And then let's just say I open up this file
• 01:53:12and I have a KO in that surface as well.
• 01:53:14So I have 3 surfaces I've made
• 01:53:17all with the same exact name in
• 01:53:20the file and I did it manually.
• 01:53:21I didn't do this through a batch,
• 01:53:23I did it manually.
• 01:53:23I'd made each one,
• 01:53:24but each one is the same surpass name.
• 01:53:27A nice little feature of Mars is
• 01:53:29the way the batch process kind of
• 01:53:31works is that it does it because
• 01:53:33it renames the new surfaces,
• 01:53:35it creates it in the same with the same name.
• 01:53:39That's how the batch processing works.
• 01:53:40If you would kind of go up here
• 01:53:41and run a batch and do that,
• 01:53:42every new surface object it makes,
• 01:53:44it's going to have the same
• 01:53:45exact name and it's going to be
• 01:53:46able to merge them together.
• 01:53:47You can do that manually based on that KO.
• 01:53:50So I can,
• 01:53:51if I select three of those
• 01:53:53guys and I do new plot up here,
• 01:53:55it's going to take those three files.
• 01:53:58Not going to do the red one,
• 01:54:00but it has the KO and in the KO down here
• 01:54:05you can go to whatever statistic
• 01:54:09doesn't matter, let's just
• 01:54:10do let's do objects for you.
• 01:54:12It doesn't really matter,
• 01:54:16right? So
• 01:54:21so I come out here and I do like
• 01:54:23a scatter plot. So the idea here,
• 01:54:26if I look at that KO surface I made,
• 01:54:29there's a column over here
• 01:54:30that says image name, right?
• 01:54:32That's the image.
• 01:54:33So this is 7/3, there's 74, there's 76.
• 01:54:36So all three images are there.
• 01:54:38All the data here from there that has
• 01:54:41that KO surface are here and it's
• 01:54:43identified based on the image name.
• 01:54:46And so from this I use Vantage
• 01:54:48to say I use Vantage a lot for
• 01:54:50kind of exporting data like this.
• 01:54:52So if you wanted to kind of just
• 01:54:54export KO from all three images
• 01:54:57and only do you know say you
• 01:54:59wanted to do so the position,
• 01:55:00the graph here is you can make a graph,
• 01:55:02right?
• 01:55:02You can do any kind of thing in Vantage.
• 01:55:03But sometimes I don't use
• 01:55:04the Vantage for graph.
• 01:55:05Sometimes I use it for data export
• 01:55:07and say I want to export area,
• 01:55:10surface area, volume and intensity, right.
• 01:55:16And then bounding box length C,
• 01:55:18whatever, right.
• 01:55:19And so I have those three statistics here,
• 01:55:21area, volume, intensity,
• 01:55:23bounding box, they're all here.
• 01:55:25And I know which one is coming from which
• 01:55:27image and I can save this data out here,
• 01:55:30right?
• 01:55:30And I can save Vantage for whatever.
• 01:55:33And so now I have all my data.
• 01:55:34It's going to be in a little bit
• 01:55:36of a not a complicated form,
• 01:55:38but it's going to be in an Excel sheet
• 01:55:40that's going to look something like this.
• 01:55:43It's not super organized,
• 01:55:45but you have the statistic here that you
• 01:55:47were really interested in those three guys,
• 01:55:49right?
• 01:55:49Your area volume,
• 01:55:50mean all the other stuff is kind of
• 01:55:53not really useful except for the one
• 01:55:57all the way over here on the right,
• 01:55:58which is your image ID,
• 01:56:01right?
• 01:56:02So this tells you what image it came from.
• 01:56:03So I would keep pretty much can
• 01:56:04get rid of everything else for
• 01:56:05the most part and kind of keep
• 01:56:06the original component and that
• 01:56:08helps you identify what it is.
• 01:56:09But basically that image ID tells
• 01:56:12you which image it came from.
• 01:56:13So that's the best way to do multiple images,
• 01:56:15multiple statistics kind of
• 01:56:17simultaneously without having to
• 01:56:19export individual data from Mrs.
• 01:56:21It's a nice little trick to kind
• 01:56:23of group them together kind of
• 01:56:24in a semi automated way.
• 01:56:26As long as you rename them
• 01:56:28exactly the same name,
• 01:56:29you can export those surfaces
• 01:56:31out together and combine.
• 01:56:33Hopefully that answers your question.
• 01:56:38When looking at object object stats,
• 01:56:40is it possible to look at the
• 01:56:42overlap area ratio to surfaces?
• 01:56:48Looking at the object object stats,
• 01:56:50is it possible to look at the
• 01:56:52overlap surface area ratio?
• 01:56:54No, we don't do a surface surface contact.
• 01:56:59I think, I think that's
• 01:57:00kind of what you're asking.
• 01:57:01There is no surface to surface contact
• 01:57:05statistic that is something to that
• 01:57:08would be a really nice statistic to have.
• 01:57:11If you reach out to me,
• 01:57:12there is an extension that kind of does
• 01:57:14a little bit of surface surface contact.
• 01:57:17It is not an officially supported
• 01:57:18bit plan extension.
• 01:57:19It's it's a custom extension
• 01:57:20I wrote a little while ago,
• 01:57:22but it's I'd be more than happy to show
• 01:57:24that to you if that's if that's what
• 01:57:25you're looking for and looking for
• 01:57:27kind of a surface surface contact value.
• 01:57:29I I can show it to people.
• 01:57:30I've used it and published it
• 01:57:31over the past couple years.
• 01:57:33That's relatively straightforward
• 01:57:34to to kind of unique application.
• 01:57:37It's not something you need
• 01:57:38the extension for.
• 01:57:38You can do it manually,
• 01:57:39but I can show you how that works to
• 01:57:41to apply that within your data set.
• 01:57:44If that's something that
• 01:57:45you're interested in.
• 01:57:45Just reach out to me separately
• 01:57:47and and set up a time.
• 01:57:48Show me your data.
• 01:57:50We can see what you're
• 01:57:51what you're trying to do,
• 01:57:53How to measure the whole stereology
• 01:57:54volume of the whole structure.
• 01:57:58Not exactly sure what you mean
• 01:58:00by that question. Specifically,
• 01:58:04if you wanted to open up a file,
• 01:58:07any file from your data,
• 01:58:10the size of your volume,
• 01:58:12the size of the acquisition data,
• 01:58:15I don't know if that's
• 01:58:16what you're looking for.
• 01:58:17If you click on volume,
• 01:58:18there are some statistics relative to
• 01:58:19the whole volume from top to bottom.
• 01:58:21So we have intensity, Max, mean,
• 01:58:23min, standard deviation sum,
• 01:58:25but you also have the
• 01:58:26volume of that acquisition.
• 01:58:27So that's the volume of your acquired
• 01:58:30image from top to bottom, right?
• 01:58:32Every single slice,
• 01:58:33every single pixel gives you that
• 01:58:35app that the volume of the entire
• 01:58:37acquisition of the data set.
• 01:58:38So the more slices you have,
• 01:58:39the bigger the volume.
• 01:58:40I don't know if that answers
• 01:58:43your question or not.
• 01:58:44If not,
• 01:58:44please reach out to me.
• 01:58:45We can we can talk more about
• 01:58:47exactly what you're what
• 01:58:48you're talking about there.
• 01:58:53OK. Where are we here? Sorry, sorry.
• 01:59:05There are some, there are some
• 01:59:08bright spots, integration at
• 01:59:10the random spots, similar sizes.
• 01:59:13That's also how do you get rid of them.
• 01:59:17I would say to answer that question,
• 01:59:19probably the machine learning
• 01:59:22probably could do that.
• 01:59:25Sometimes the machine learning,
• 01:59:27object based machine learning
• 01:59:28after you generate your objects,
• 01:59:30being able to kind of say hey that's a cell,
• 01:59:31that's a cell, that's noise,
• 01:59:33that's noise you'd be crazy to to to
• 01:59:36think sometimes it's not going to work,
• 01:59:38but sometimes I'm like that's
• 01:59:39not going to work and it actually
• 01:59:41worked to me without seeing the
• 01:59:43data that's probably going to be
• 01:59:45your best bet to to get rid of them.
• 01:59:47And again even the pixel classification
• 01:59:49a lot of times pixel classification is
• 01:59:51also good enough to get rid of them.
• 01:59:54Point in case I didn't,
• 01:59:56I was going to get into it but I didn't.
• 01:59:59I wanted to show you this example
• 02:00:00because it kind of demonstrates
• 02:00:02the power of the pixel classifier.
• 02:00:03Let's see if I put it in here.
• 02:00:06I think I put it in.
• 02:00:09Yeah, here. So I'm not going to
• 02:00:11go through the process here,
• 02:00:12but I did this right before we came in here.
• 02:00:15So this is kind of a one
• 02:00:17of our demo data sets.
• 02:00:19You can see there's a lot
• 02:00:20of blue in this data set.
• 02:00:21There's a lot of small little cells
• 02:00:23over here and then there's these
• 02:00:25big nurse nuclei in the middle here
• 02:00:27that create these big surfaces,
• 02:00:29if you can make those big surfaces, right.
• 02:00:31However, if I were to make a regular
• 02:00:35surface in a Morris and I do the
• 02:00:37pixel classification to try to
• 02:00:39render these small guys out here,
• 02:00:41you're going to get surfaces
• 02:00:42that are going to create.
• 02:00:44You're going to get surfaces that
• 02:00:45are going to create everything.
• 02:00:47You're going to get surfaces out here,
• 02:00:49you're going to get sub segmentation here,
• 02:00:51you get the surface out here,
• 02:00:52you get a lot of junk everywhere.
• 02:00:53However, if you do pixel classification
• 02:00:56and I do something like this,
• 02:00:57I can train it so that I'm only
• 02:01:00rendering these pixels,
• 02:01:01these the nuclei that are out here.
• 02:01:03It's not rendering these
• 02:01:05Big Blue nurse nuclei here.
• 02:01:08Why?
• 02:01:08Because I trained it as background,
• 02:01:10right?
• 02:01:10And if I come up here and I just
• 02:01:11show you real quickly what I did,
• 02:01:13I think it'll come up.
• 02:01:17You come in here, you can see I trained.
• 02:01:22I trained that as background, right?
• 02:01:24I came in here and you can see where is it.
• 02:01:32You know, I trained all these. I
• 02:01:36guess that's it. Oh, there.
• 02:01:39Yeah. So I trained all these
• 02:01:41pixels here as background,
• 02:01:42all that as background,
• 02:01:43all that as background.
• 02:01:44And then I went up here and I
• 02:01:45trained that as signal, right.
• 02:01:46And so even though it's blue and it's,
• 02:01:48it's a little bit brighter and blue,
• 02:01:50but it's still blue.
• 02:01:51I'm able to train that as background
• 02:01:53and totally ignore it for the creation.
• 02:01:56Now that's that might be an option as well.
• 02:01:58Again, the pixel classifier kind
• 02:02:00of surprises me sometime how well
• 02:02:01it's able to kind of separate
• 02:02:03separate those spots there.
• 02:02:04But again,
• 02:02:04without seeing the data set it'd be hard.
• 02:02:06So again, reach out to support.
• 02:02:08We can take a look at it together,
• 02:02:11such as one embryo which has many cells.
• 02:02:13I would like to measure how big size
• 02:02:16of the volume of the whole embryo,
• 02:02:19not only one individual
• 02:02:21cell volume of the embryo.
• 02:02:23Again, I would say in a case like that,
• 02:02:25probably the pixel classification
• 02:02:27might work really well.
• 02:02:29I've used that on a number
• 02:02:30of occasions to kind of say,
• 02:02:32you know, what if I were to open
• 02:02:35up a data set like here and say,
• 02:02:39you know what, I want to measure
• 02:02:41the size of this whole structure,
• 02:02:42not just the cells, not just the cells.
• 02:02:45I want to measure the whole cell,
• 02:02:46including all the intercellular
• 02:02:48space essentially.
• 02:02:50So I would, you know again come up here.
• 02:02:53I'd probably do a little bit of
• 02:02:54smoothing background strand,
• 02:02:55I'd probably do all channels come in here.
• 02:02:58I can train it and say you know what I'm
• 02:03:00going to make it pretty broad here and I'll,
• 02:03:02you know,
• 02:03:02I come up here and is it going to let me,
• 02:03:04it's not going to let me.
• 02:03:05Sorry, didn't try to reopen,
• 02:03:07sorry.
• 02:03:07So
• 02:03:13I can come up here and I
• 02:03:15would just train it to say
• 02:03:23takes it here,
• 02:03:24it's going to let me draw oh sorry,
• 02:03:26I'm on the wrong button.
• 02:03:27So yeah I do that all the time.
• 02:03:28Make sure you're on the the,
• 02:03:30not on the circle select,
• 02:03:31but on the on the thing here.
• 02:03:33But anyway, you can come up here.
• 02:03:34You're going to train this as a
• 02:03:36signal and you come up here and you
• 02:03:38can train this as background, right?
• 02:03:39And you get you'll get a
• 02:03:41whole a whole oops sorry,
• 02:03:46background. Train it as
• 02:03:48background and you can get.
• 02:03:51Hopefully if you train it right,
• 02:03:53you'll get one,
• 02:03:55one big BLOB including everything.
• 02:03:59And again, a lot of times there's
• 02:04:01auto fluorescence and things
• 02:04:01like that that get picked up.
• 02:04:03And clearly background is background
• 02:04:04and you'll get something that
• 02:04:06looks like this that'll be able
• 02:04:07to kind of give you that idea of
• 02:04:08that one big volume structure.
• 02:04:10And then then you can just go here
• 02:04:12and look at total volume down here
• 02:04:14and that'll give you the shape,
• 02:04:15total size of the the embryo.
• 02:04:17Again just depends on the data set,
• 02:04:19but that that is pretty easy way of
• 02:04:21doing it through pixel classification.
• 02:04:23You might be able to even do it
• 02:04:24through just intensity based.
• 02:04:25Again a lot of times auto fluorescence
• 02:04:28can be your friend in terms of
• 02:04:30rendering kind of a whole cell object.
• 02:04:33I use that on many occasions and say,
• 02:04:34hey, you know what the auto
• 02:04:35fluorescence in that one channel,
• 02:04:36even though I don't think it's
• 02:04:38good for this segmentation,
• 02:04:39I can use it to kind of render the whole
• 02:04:41cell that's I've used that on many,
• 02:04:43many different occasions.
• 02:04:45OK And last one here.
• 02:04:48Is there a way of visualizing
• 02:04:49shells in a nucleus as the distance
• 02:04:51from the center of the nucleus to
• 02:04:53how far is this are away from the
• 02:04:54center or the edge of a nucleus?
• 02:04:56Oh gosh,
• 02:04:57you are jumping into a totally
• 02:04:59module that I didn't get into today.
• 02:05:02So the short answer is yes,
• 02:05:04we can do that,
• 02:05:05but it is using a module called Amar cell.
• 02:05:08It's this guy up here where you would
• 02:05:11actually do a detection of the whole cell,
• 02:05:15you would do a detection of the nuclei
• 02:05:18and then you can measure objects
• 02:05:21relative to that nuclei and measure
• 02:05:23how close they are to the center,
• 02:05:25how close they are to the membrane.
• 02:05:27And there's a lot of different
• 02:05:29parameters there that we can measure
• 02:05:30relative to the edge of the nucleus,
• 02:05:33but it's you can do it separately
• 02:05:36I suppose as well. But the Mr.
• 02:05:38cell module is, is the way to go basically.
• 02:05:41I can show you really quickly
• 02:05:43I don't want to get into it.
• 02:05:44But again,
• 02:05:44I think if you have a specific
• 02:05:46question there,
• 02:05:46again I can cover everything
• 02:05:47in a lot of detail today.
• 02:05:49Let me see if I can find dumb,
• 02:05:53dumb example here.
• 02:05:57This guy here, I guess, right?
• 02:06:01So this is a cell.
• 02:06:03Let's just say come in here.
• 02:06:05I think I can build this
• 02:06:07really quick. See here.
• 02:06:08See this guy here? So we do.
• 02:06:13The nuclei is blue.
• 02:06:15You come in here, you're
• 02:06:17going to segment the nuclei.
• 02:06:20It's not working.
• 02:06:27Hold on, let me start over.
• 02:06:29I'm not sure why it's not working.
• 02:06:30It's not showing up in the threshold.
• 02:06:37There
• 02:06:42we go, Right. So there's our nuclei.
• 02:06:47One big nuclei.
• 02:06:49We have our cell for big cell labeling.
• 02:06:53Here again, I'm not going to do
• 02:06:55super duper fancy here, right?
• 02:06:57So there's our cell and then we
• 02:07:00have some vesicles in the cells.
• 02:07:02So I come in here, turn this guy off,
• 02:07:06We can detect, we can detect spots.
• 02:07:08That's probably the easiest way to do it.
• 02:07:10So you have a whole bunch of spots
• 02:07:12within the structure and then you
• 02:07:14have a whole ton of statistics.
• 02:07:15Now again, you have a cell option.
• 02:07:17Now the visualization of the Marcel
• 02:07:19module is. It should be desired.
• 02:07:20It's not quite the same
• 02:07:21as the surface rendering,
• 02:07:22but the statistics are all there.
• 02:07:24So there's the cell, there's your vesicles,
• 02:07:26there's your nuclei.
• 02:07:27All the statistics are here.
• 02:07:29You can look at these statistics on the Mr.
• 02:07:31cell feature.
• 02:07:32So for example, cell
• 02:07:37vesicle. Where's the vesicle? Vesicle.
• 02:07:41Distance? Maybe I haven't turned on.
• 02:07:48Yeah, I can turn it on,
• 02:07:48but there's a lot of statistics here.
• 02:07:51So that's fulls. So distance.
• 02:07:54Here it is. Yeah.
• 02:07:55So distance to cell, membrane distance,
• 02:07:57the closest nucleus distance, nucleus center.
• 02:08:00So there's a lot of different parameters
• 02:08:02here that you can export out and
• 02:08:05get some cell to cell statistics.
• 02:08:07Now the challenge here is to do
• 02:08:09the rendering inside of a Marcel.
• 02:08:10It works really well.
• 02:08:12You can also import surfaces
• 02:08:13into a Marcel there.
• 02:08:14There's an import process here.
• 02:08:15So if you did your segmentation in surfaces,
• 02:08:18you can import them into a Marcel.
• 02:08:19Sometimes that's not super efficient,
• 02:08:21but it can get definitely get the job
• 02:08:23done and get you those statistics that
• 02:08:25are automatically calculated for you.
• 02:08:27So I think I'm going to stop there.
• 02:08:29I know we went really long
• 02:08:31here and we got a late start,
• 02:08:32but I think I got to all the questions.
• 02:08:34I think my my biggest and I appreciate you
• 02:08:37all kind of sticking with me to the end here.
• 02:08:40As I said before,
• 02:08:41you guys have a maintenance contract
• 02:08:43with the core facility gives you
• 02:08:45access to myself and our support team.
• 02:08:48Whenever you have a question on using the
• 02:08:50software with a particular application,
• 02:08:52I'll be more than happy to
• 02:08:53kind of ask for that data,
• 02:08:54walk you through it one-on-one.
• 02:08:56Please just reach out to send
• 02:08:58an e-mail to your support,
• 02:08:59tell me who you are.
• 02:09:00Tell me who's license you're using
• 02:09:02and things like that with you.
• 02:09:04Any any of you want any more information
• 02:09:07on the modules that I didn't cover,
• 02:09:09Again, please reach out.
• 02:09:10I think you guys have all of
• 02:09:12the modules available to you,
• 02:09:14so there's a lot of different options
• 02:09:15out there to to analyze your data.
• 02:09:17And I'm more than happy to have
• 02:09:19conversations and teach you some of the
• 02:09:21things that hey maybe you can do this,
• 02:09:22maybe you can do that and try to get
• 02:09:25maybe some things that you didn't think
• 02:09:26about for your for your data analysis.
• 02:09:30So I appreciate your time and effort.
• 02:09:33I know this is a a long meeting but
• 02:09:35hopefully it gives you an idea of of
• 02:09:36what you can do with the software.
• 02:09:38When you go down and sit in front
• 02:09:40of it you you'll have a a thing you
• 02:09:41will get a video automatically I
• 02:09:43think at the end of this meeting.
• 02:09:45I don't know how long it takes
• 02:09:46to get that video sent,
• 02:09:47but if you signed up and registered
• 02:09:48you should get an automatic
• 02:09:50video recording of the session.
• 02:09:51If you don't please let me know.
• 02:09:52I think I'm going to get one too.
• 02:09:54So hopefully if I get it,
• 02:09:55you guys will get it as well.
• 02:09:57But I'll have the recording
• 02:09:59available if if you don't,
• 02:10:01if you don't get that link to
• 02:10:02the to the video. All right.
• 02:10:04Well, I hope that is what you
• 02:10:06were hoping for, Matthias.
• 02:10:07Yeah, sure.
• 02:10:10Hopefully it gives you guys a a good
• 02:10:12starting point for a lot of different
• 02:10:15tools out there to to use our software.
• 02:10:17All right.
• 02:10:18So I guess we should just all thanks
• 02:10:21Matthew for dedicating so much time
• 02:10:23and covering so much material and
• 02:10:25it's great that we we have this
• 02:10:28starting point and as Matthew said,
• 02:10:30we have a maintenance contract,
• 02:10:31you guys can always reach out
• 02:10:33to the supporter.
• 02:10:34Demarius just mentioned that the license
• 02:10:36is CCMI and it's under my name I believe.
• 02:10:40So he knows how to to route
• 02:10:42that and and as I said,
• 02:10:44he's going to,
• 02:10:45you're going to get a recording.
• 02:10:46And we hope that this has been very
• 02:10:49useful to all of you as it has been to me.
• 02:10:52And I'm looking forward to working
• 02:10:54more with you and with more
• 02:10:56advance from Marty's features.
• 02:10:58Yeah.
• 02:10:58Yeah.
• 02:10:58You're more than welcome to
• 02:10:59have sessions with me as well.
• 02:11:00If you want to learn anything
• 02:11:01about what we covered more detail
• 02:11:03and get more technical, we can.
• 02:11:04We can do that as well. Yeah.
• 02:11:06Great.
• 02:11:06Great.
• 02:11:07So, yeah, everybody, thank you very much.
• 02:11:10And Matthew, I wanted to chat with you
• 02:11:12a couple minutes if you have time,
• 02:11:13but I don't know if you want to call me.
• 02:11:17Do you mind calling me? Oh, yeah.
• 02:11:19I can give you a call. Yeah.
• 02:11:20OK Let me just send you.
• 02:11:22Let me just send you the number here.
• 02:11:29OK. All right, guys. Yeah.
• 02:11:32Thank you. I don't know, Matthew,
• 02:11:33if you can end the session.
• 02:11:35I end the session. So, yeah, I'll end.
• 02:11:37I'll end the session.
• 02:11:38Yeah, I'm going to end it. OK.
• 02:11:39Yeah. I just send you on the chat,
• 02:11:40the phone number.
• 02:11:43Where's the chat?
• 02:11:43Let me make sure I can see it.
• 02:11:47OK. Make sure I can copy that out.
• 02:11:52OK. All right. I'll call you shortly.
• 02:11:54OK. All right. Thank you, everyone.
• 02:11:56Have a great day. You too. Bye.