Research

The majority of the literature about AI studies in neuro-oncology uses databases that consist of about 200 patients, with the largest databases being close to 500 patients. As a team, we have identified a database of gliomas from the Yale Tumor Registry that includes 1,991 gliomas with available outcomes of survival and molecular profile.

Our research focuses on the development of auto-segmentation and feature extraction tools in neuro-oncological MRI. As a team, we have developed a nnU-Net, a deep learning-based segmentation method for gliomas and made this algorithm available on clinical Visage PACS interface. This allows the direct implementation of AI tools into clinical practice for ease of use, improvement of workflow and patient outcome, such as time to diagnosis, choice of treatment approaches or overall survival.

ImagineQuant’s preclinical division is dedicated to generating a deeper understanding of the neural and psychological effects of radiation exposure during brain cancer treatment. Our studies of rodent models employ various in vivo and in vitro analytical techniques, such as SV2A PET and autoradiography, to quantify changes in synaptic density and further elucidate the pathomechanisms of cognitive decline.

An important clinical problem today is selection of patients for treatment with immune checkpoint inhibitor (ICI) therapy, which is effective in only about 20% of patients with metastatic disease. Currently, one of the biomarkers that predicts patient response to ICI, is immunohistochemistry-based measurement of PD-L1 levels, which is limited due to invasive nature of biopsy, inability to biopsy multiple lesions in patients with widely metastatic disease, and sampling error. We propose to develop a protocol for imaging based quantitative measurement of PD-L1 using a novel PET tracer and to validate this method against immunohistochemistry in resected primary and metastatic lesions.