Skip to Main Content

Yale Pathologists Say AI, Algorithms in Digital Pathology Will Have Lasting Impact on Education, Clinical Practice

September 02, 2022

The continued emergence and development of artificial intelligence and algorithms in digital pathology will greatly impact the practice and teaching of pathology, Angelique W. Levi, MD, Associate Professor of Pathology, and Sudhir Perincheri, MBBS, PhD, Assistant Professor of Pathology, said in a recent podcast.

Speaking with Dr. Joseph Anderson on his Digital Pathology Today podcast, Dr. Levi, Vice Chair and Director of Pathology Outreach Services, and Dr. Perincheri, Director of Digital Pathology, said the technology will have a lasting influence on education and clinical practice, especially in community settings.

AI-based tools and algorithms are emerging in digital pathology and can help diagnose cases or at least provide a second review to the pathologist’s work. But as Dr. Anderson noted, before we can embrace this technology, we must ensure these tools are developed carefully.

“How do we know we have a robust platform that we can build on?” Dr. Anderson asked.

It all depends on how well the algorithm is written, Dr. Perincheri said.

“If you look at the algorithms that have been in the news the most, they usually come out of big centers trained on an extensive number of cases. I think the most important thing is how extensive is the dataset, how good is it, and how comprehensively has it been annotated to train the AI algorithm,” he said. “Studies have shown that these algorithms are able to mine features that the human eye simply does not see. But to do that, it is incredibly important to have a dataset where you know outcomes. It ultimately boils down to the robustness of the data spanning many number of years so that you can actually tie features of diagnosis to long-term outcomes.”

Dr. Perincheri said the technology could have a more practical impact in a community setting as opposed to an academic practice, where cases are reviewed by several sets of well-trained eyes.

“If you move it away from an academic setting, where there are a number of people looking at the same individual case, to more of a community setting or a private practice setting, where, perhaps, not the same number of eyes are reviewing a case, then the utility of these algorithms should show up,” he said. “In fact, there are papers published now that show the detection of cancer in various areas goes up when these tools are applied as a diagnostic.”

“Considerations of turn-around time in community practice are relevant,” Dr. Levi said. “And there’s the Quality Control component – hospitals use pathology and pathologists to guide patient safety and safety data – so that component is very useful in that setting. Improving workflow in a busy practice can also hopefully assist in preventing burnout and improve overall well-being. Those are some features that I feel, in community practice, these AI tools would be very useful because these practices don’t have the access an academic practice does.”

Dr. Perincheri said the technology will improve accuracy and can also be used as a second test.

“The quantitative measures that form part of our reports are going to be much more accurate,” he said. “We’ll be moving from manual measurements to automated measurements on a digital image, which are going to be, in my opinion, far more accurate. And I can envision a scenario where every negative case that I sign out and say there’s no cancer here, I can put it through the algorithm as a second test to make sure that I have not made an error.”

How the tools are used and in what setting is extremely important, Dr. Levi said.

“If you’re in a practice where prostate cancer isn’t as common as in an academic tertiary center, do you need a very sensitive AL algorithm that’s going to pick up all of the atypical small proliferations? It really is a tool that we have to apply carefully and use so that in each setting, it doesn’t do the opposite of what we intended it to do,” she said.

They agreed the technology will have an undeniable influence on education.

“As an academic program, I think it’s incumbent on us to put the mindset in our trainees, give them the exposure to these technologies, and emphasize some of these technologies in their training so that as they join in, they can take this field into the future,” Dr. Perincheri said. “Because if we don’t, the data scientists will. If we want to take the ownership of this digital revolution, it’s incumbent on us to make sure our trainees are comfortable with these aspects of pathology.”

“We will always need to learn the fundamentals of histopathology and morphology . . . but the new pieces are the ability to expose residents to this technology and give them the opportunity to innovate,” Dr. Levi said. “Mentorship and the exposure to this during the educational process is really the investment in creating pathologists who are going to be driven to lead in this field.”

As for the future, Dr. Levi envisions a day when the technology will be everywhere.

“Part of what excites me is not just improving the standard of care here at Yale but having that extend into communities beyond academic institutions to increase the access to this technology in a way that can have the greatest impact, and extend this care perhaps in other countries where there isn’t subspecialty expertise or training – or even a pathologist,” she said.

“What we need to do is get out of our silos and start emphasizing the integrative aspect of diagnostic medicine,” Dr. Perincheri added. “I think if we do that, we will continue to be a vital cog in the healthcare space and add value to that space.”

Submitted by Terence P. Corcoran on September 02, 2022