Accomplishments

Formed and lead an international collaboration (Imperial College UK, Mayo, NIH, Moffitt) to develop a deep learning system for analysis of clinical-grade MRI exams of brain cancer patients. Created a pipeline with multiple 3D deep neural networks to: enhance the MR exams; segment the brain even in the presence of major pathology; segment brain tumors; label anatomical regions; and, extract image signatures for predictive analytics. System runs on AWS in 1 - 2 mins. per MR exam. Recruited 20 neuroradiologists to perform the first randomized blinded comparison of human and DNN segmentation (AWS AppStream 2.0). They judged the DNN results to be better, on average, than those from the technicians who created the DNN training data.

Designed and implemented a DNN pipeline (caBERTnet) to identify and extract tumor site and histology from pathology reports. The system predicts fine-grained ICD-O-3 codes across 214 tumor sites and 193 histologies with top-5 accuracies of 94% and 95%, respectively. (Nvidia GTC '21 talk & demo)

Recruited, supervised, and mentored a dynamic team including 2 PhD scientists, 3 post-docs, 4 PhD students, and multiple MSc & BSc students. Fostered and maintained key relationships with Arizona State University departments of CS, Industrial Eng. and Biomedical Informatics. Fostered and maintained key collaborations with Mayo Clinic departments of Radiology, Neurology, Neurosurgery, Oncology, Hematology, Cardiology, Dermatology, Preventive Medicine, Bioinformatics, and, Information Technology.

Spearheaded efforts to configure, install, and support ResolutionMD Mobile backend and frontend hardware and software for Mayo Clinic AZ. Improved precision health via rapid, simple, and secure access to medical images and reports, anywhere. Project success prompted Mayo to procure Enterprise mobile image viewing technology.