SEMINAR: Applying Bayesian Deep Neural Networks to Neuroimaging

Patrick McClure, Ph.D.
When Feb 20, 2019
from 12:00 PM to 01:30 PM
Where Shumaker Research Bldg, RM 139
Contact Name
Contact Phone 502-852-7485
Add event to calendarvCal
iCal

Abstract: 

Deep neural networks (DNNs) are being applied to increasingly diverse areas, including medical image analysis. One area of active research is improving the estimation of DNN uncertainty. Better uncertainty estimation leads to better performance for a variety of problems, including transfer learning, distributed learning, and anomaly detection. These issues are particularly common for neuroimaging. Bayesian DNNs have been proposed as a principled solution for improving DNN uncertainty estimation by learning distributions of DNNs. In this talk, we will briefly introduce Bayesian DNNs and discuss multiple uses of uncertainty. We will then demonstrate the usefulness of Bayesian DNNs in neuroimaging, using MRI brain segmentation as a case study.

Speaker:  

Patrick McClure, PhD received his Bachelor’s in Bioengineering and his Master’s in Computer Science from the University of Louisville, while conducting medical imaging research under Dr. Ayman El-Baz. He then received a PhD in Computational Neuroscience from the University of Cambridge. While at Cambridge, he researched deep neural network models of human visual perception and decision making under the supervision of Dr. Nikolaus Kriegeskorte. Dr. McClure is currently a Machine Learning Research Scientist at the NIH, where he uses and develops deep learning methods for both computational neuroscience and neuroimaging.

 

Seminar Video on Facebook