Scientific Lectures //
Deep Learning Models for Brain Imaging: Model Depth Enhances Discovery Power
Sergey Plis, PhD.- Assistant Professor of Translational Neuroscience & Director of Machine Learning in Neuroscience Lab; The Mind Research Network
Presented: June 17, 2014
Deep learning methods are breaking records in the areas of speech, signal, image, video and text mining and recognition by improving state of the art classification accuracy by, sometimes, more than 30% where the prior decade struggled to obtain a 1-2% improvement (Krizhevsky 2012, Le2012). These are very important tasks for brain imaging and neuroscience discovery. In this work we demonstrate our results in application of deep learning methods to brain imaging data from schizophrenia and Huntington disease studies. Our results show: automatic feature learning and model depth enable detection of latent relations in neuroimaging data and improve discriminative power.