Meeting Banner
Abstract #3198

ProxVNET: A proximal gradient descent-based deep learning model for dipole inversion in susceptibility mapping

Christian Kames1,2, Jonathan Doucette1,2, and Alexander Rauscher1,2,3
1Department of Physics and Astronomy, University of British Columbia, Vancouver, BC, Canada, 2UBC MRI Research Centre, University of British Columbia, Vancouver, BC, Canada, 3Department of Pediatrics, University of British Columbia, Vancouver, BC, Canada

A deep learning model, ProxVNET, is proposed to solve the ill-posed dipole inversion in susceptibily mapping. ProxVNET is derived from unrolled proximal gradient descent iterations wherein the proximal operator is implemented as a V-Net and is itself learned. ProxVNET is shown to outperform the U-Net-based dipole inversion deep learning model QSMnet when compared to COSMOS reconstructed susceptibility maps.

This abstract and the presentation materials are available to 2020 meeting attendees and eLibrary customers only; a login is required.

Join Here