Artificial intelligence for pancreatic cancer diagnosis | Shayan Monabbati
Interview with Shayan Monabbati
How would you explain your research outcomes to the non-scientific community?
We were able to create an automated machine learning-based pipeline that can identify malignancies from high resolution images of patient biopsies from the bile duct, which is one of the rarest and deadliest forms of cancer. Our model is able to diagnose the most difficult cases without any false positives, which are often misdiagnosed and lead to unnecessary treatments for patients. We were able to improve the true positive rate from around 44% to 68% when compared to cytopathologist performance.
How do these findings contribute to your research area?
This research is primarily designed to serve as a clinical support tool for cytopathologists, especially those who specialize in the biliary and pancreatobiliary tracts. One of the most difficult challenges in this field is to diagnose patients with atypical cell morphologies, which have extremely subjective criteria for malignancy amongst experts. Our work will aid in being able to further stratify those cases into a definitive diagnosis so that the clinician can prescribe the best treatment available for the patient.
Read full interview here: