Machine Learning & Radiomics for Imaging Biomarker Discovery
AI and Medical Imaging for Better Care
Medical imaging relies almost exclusively on human interpretation of CT or MRI to guide clinical decision making. We leverage recent development in the field of artificial intelligence to mine digital data in medical images to discover biomarkers related to cancer aggressiveness and therapeutic targets.
-
nnUnet has been integrated into our pipelines for rapid tumor segmentation
Output of nnUnet trained models in prostate cancer, kidney cancer, liver cancer and pancreatic cancer (clokwise)
-
A Step-by-Step Tutorial for Radiologists on Building a Neural Network for Segmentation
e-Poster at SAR 2023 in Austin Texas
-
Externally Validated RadScore for Pancreatic Cancer Published
Gerard Healy with Emmanuel Salinas‑Miranda have published an externally validated RadScore (Radiomics Score) showing prognostic predictive ability in pancreatic cancer preoperatively, prior to the availability of tissue and surgical staging. This may have value in selecting patients for neoadjuvant therapies.
-
AI for COVID-19 - The EXAM Study is Published
20 hospitals, including the Sinai Health System at the University of Toronto, collaborated on a Federated Learning AI model led by NVIDIA and Mass General Brigham to triage COVID19 patients based on CXR and clinical parameters on admission to the Emergency Room. Our lab in collaboration with members of the Dept. of Emergency Medicine, Internal Medicine and Laboratory Medicine compiled the data and set up the federated learning infrastructure based on the NVidia Clara Platform. Learn more about the initiative, published today in Nature Medicine.
Watch video