Abstract #4780

Neural style transfer: Applications in cardiac MR image registration

Alper Ozan Turgut¹, Matthew Van Houten², Junyu Wang², Xue Feng², and Michael Salerno³

¹School of Medicine, University of Virginia, Charlottesville, VA, United States, ²Biomedical Engineering, University of Virginia, Charlottesville, VA, United States, ³Department of Medicine and Radiology, Stanford University, Stanford, CA, United States

Synopsis

Contrast-enhanced cardiac magnetic resonance (CMR) stress perfusion imaging shows excellent utility in evaluating coronary artery disease¹. Registering perfusion CMR image series is difficult due to the varying image contrast. Neural style transfer is a deep learning method used to transfer the “style” of one domain to another while preserving the content. Two neural style transfer networks were implemented in Python using TensorFlow and PyTorch. Training of each network was done using three, slice matched patient profiles and cine-like perfusion images were generated and registered. This method is compared to a KL-transform based registration approach.

This abstract and the presentation materials are available to members only; a login is required.

Join Here