Deep Learning for Multimodal and Dynamic Functional Neuroimaging
Speaker: Archana Venkataraman, PhD
John C. Malone Assistant Professor, John Hopkins University, USA
Talk abstract
Deep learning has disrupted nearly every major field of study from computer vision to genomics. The unrivaled power of these models offers a new and exciting way to study the intricacies of the brain via functional neuroimaging. On the flipside, deep networks lack interpretability and are notorious for overfitting on smaller datasets. My lab tackles these challenges through careful problem formulation, architectural design, and model training. This talk will highlight two ongoing projects that demonstrate both the scientific and translational potential of deep learning for fMRI. In the first project, we develop a novel deep network architecture to integrate imaging and genetics data, as guided by diagnosis. Our model consists of a coupled autoencoder and classifier. The encoder learns a non-linear subspace shared between the input data modalities. The classifier and the decoder act as regularizers to ensure that the low-dimensional space captures predictive differences between patients and controls. We use a learnable dropout layer to extract interpretable biomarkers from the data, and our unique training strategy can easily accommodate missing data modalities across subjects. In the second project, we develop an end-to-end deep learning framework that uses dynamic functional connectivity to simultaneously localize the language and motor areas of the eloquent cortex in brain tumor patients. Our model leverages specialized convolutional filters that extract graph-based features from the dynamic connectivity matrices, an LSTM attention network to weigh the relevant time points and multitask classification to simultaneously localize different eloquent subsystems.
The Feindel Virtual Brain and Mind (VBM) Seminar Series
Advancing the vision of Dr. William Feindel (1918–2014), former Director of the Neuro (1972–1984) and Founding Director of the McConnell Brain Imaging Centre (1984), to constantly bridge the clinical and research realms. The talks will highlight the latest advances and discoveries in neuropsychology, cognitive neuroscience, and neuroimaging.
Speakers will include scientists from across The Neuro, as well as colleagues and collaborators locally and from around the world. The series is intended to provide a virtual forum for scientists and trainees to continue to foster interdisciplinary exchanges on the mechanisms, diagnosis and treatment of brain and cognitive disorders. If you, or someone you know, would be interested in presenting please contact us to secure a spot on the schedule.
To register: Eventbrite
