Time and Location
Oct. 12th, 04:30 PM to 05:10 PM (PDT- Pacific Daylight Time)
Meeting Room 14, Vancouver Convention Center East Building Level 1
Abstract
For over a decade, our ENIGMA Consortium has coordinated the largest worldwide neuroimaging studies of over 30 brain diseases, ranging from neurology (Parkinson’s disease, epilepsy, and brain trauma), to psychiatry (PTSD, bipolar disorder, schizophrenia, depression and addiction) to brain development (autism, OCD, and ADHD). These studies now merge MRI, DTI, fMRI, EEG and MEG with genome-wide, epigenetic, geocoded, and clinical data from over 100,000 people across 45 countries. AI and deep learning methods can benefit from these massive worldwide datasets to improve diagnosis, prognosis, and treatment selection for patients with these disorders, and to discover new drug targets and treatable risk factors. Here we review several categories of AI methods that are now accelerating large-scale studies of disease, including pretrained CNNs, Vision Transformers, GANs, diffusion models, and federated learning. We explain how to pretrain these models to maximize efficiency, comparing popular computer vision architectures that are being adapted to 3D and 4D medical imaging, as well as generative models that create synthetic brain MRIs, and multimodal variational autoencoders. We showcase some N-site deep learning challenges, including neural style transfer to handle domain shift, and some new directions including image enhancement, cross-modal image synthesis, and methods that work on multi-modal data.
Speaker’s Bio
Dr. Paul Thompson is a Professor of Neurology, Psychiatry, Radiology, Pediatrics, Engineering, and Ophthalmology in the Keck School of Medicine of USC, and Director of the ENIGMA Center for Worldwide Medicine, Imaging & Genomics - a $11M NIH Center of Excellence in Big Data Computing. Dr. Thompson directs the ENIGMA Consortium, a global alliance of 307 scientists in 33 countries who conduct the largest studies of 10 major brain diseases – ranging from schizophrenia, depression, ADHD, bipolar illness and OCD, to HIV and addictions on the brain. Collaborating with imaging labs around the world, Dr. Thompson and his students have published over 1,300 publications (h-index: 116) describing novel mathematical and computational strategies for analyzing brain image databases, detecting pathology in individual patients and groups, and creating disease-specific atlases of the human brain.