December 3, 2019

12:00 pm / 1:30 pm

Venue

Clark Hall, Room 316

12:00 – 1:00 Seminar
1:00 – 1:30 Lunch

Abstract:  Machine learning formedical image analysis provides huge opportunities for advances in healthcare,such assisting in patient diagnosis and permitting advances in personalizedmedicine. However, several problems unique to medical image analysis, includingthe prevalence of uncertainties, challenge popular machine learning methodsdeveloped for computer vision applications.  This talk will focus on theparticular contexts of brain tumour/lesion detection and segmentation inpatient images, where probabilistic graphical models have been successfullyapplied to large-scale, multi-scanner, multi-center clinical trial datasets ofpatients with Multiple Sclerosis as well as to the MICCAI BRaTs brain tumoursegmentation challenge datasets. We will describe recent work exploringmeasures of uncertainty in deeplearning lesion detection and segmentationmodels, and illustrate how propagating uncertainties across cascaded medical imagingtasks can improve deeplearning inference. Finally, current work on predictionof future lesion activity and disease progression based on baseline MRI will bebriefly described.

 

Bio: Tal Arbel is a Professor in the Department of Electrical andComputer Engineering, and the Director of the Probabilistic Vision Group andMedical Imaging Lab in the Centre for Intelligent Machines, McGill University,and an associate member of MILA, the Quebec Artificial Intelligence Institute.Prof. Arbel’s research focuses on development of probabilistic machine learningmethods in computer vision for medical image analysis, with a wide range ofapplications in neurology and neurosurgery. Tools developed in her lab forlesion detection and segmentation have been integrated into the softwareanalysis pipeline of an industrial partner for usage in clinical trial drugdevelopment, where the methods have assisted in the analysis of almost all thenew MS treatments currently being used worldwide. Prof. Arbel has authored over100 peer-reviewed papers and wasthe 2019 recipient of the McGill EngineeringChristophe Pierre Research Award. She has co-organized a number of majorinternational conferences in bothfields, including serving as co-organizer andsatellite events chair for MICCAI 2017, area chair for ICCV, CVPR, MICCAI, MIDLand General Chair for a joint national conference (AI/GI/CRV/IS).  She iscurrently an Associate Editor (AE) for IEEE Transactions on Pattern Analysisand Machine Intelligence (TPAMI) and was an AE for the Journal of ComputerVision and ImageUnderstanding (CVIU).