Pritam Mukherjee, PhD

My research into the application of machine learning and AI in pediatric radiology focuses on the areas listed below. By concentrating on these topics, I help advance not only what is possible within radiology but how we look at and use patient data throughout a patient’s healthcare experience: at diagnosis, during treatment and into survivorship.

A main area of my research explores AI methodologies to automate the interpretation of radiology images across various imaging modalities, such as MR and CT. By automating the detection, segmentation, and follow-up of abnormalities shown on radiologic images over time, we can speed up and improve diagnostic capabilities in pediatric radiology practices.  

Large language models (LLMs) represent a paradigm shift in how we process information, and my research aims to better understand how they can be used in clinical applications. I am most interested in exploring how LLMs can be utilized as a tool to help with data integration in clinical settings, from labelling and extracting information from radiology reports to answering radiology questions and generating reports.

Patients diagnosed with childhood cancer and other catastrophic diseases will acquire a vast amount of health information as they move through their care journey. All this information, ranging from various medical imaging scans to genomic sequencing, is stored in an electronic health record. Before making clinical decisions that will impact the patient’s outcomes, clinicians must review all this data, subsets of health information that can span years, if not decades. In this research area, my goal is to use AI to integrate the vast array of clinical data to help clinicians make faster and more informed decisions for their patients.

While my current work spans tool development, disease trajectory modeling, and quantitative phenotyping, my future goal is to advance the creation and use of a patient “digital twin” — a simulation of a patient and their condition, which can be modified to simulate how they would respond to potential treatment or disease progression. As we learn how best to develop and implement digital twins in clinical settings, the underlying hope is that they can aid in predictive modeling, prognosis, and even more informed clinical decision making.

All areas of my research support and touch one another as we work to discern how best to use AI and machine learning to advance cures and optimal outcomes for patients with childhood cancer and other catastrophic diseases.