Meeting Banner
Abstract #1253

Deep Learning based pseudo-CT computation and its application for PET/MR attenuation correction and MR-guided radiation therapy planning

Sandeep Kaushik1, Cristina Cozzini2, Mikael Bylund 3, Jaewon Yang4, Dattesh Shanbhag1, Joakim Jonsson3, Josef Lundman3, Thomas Hope4, Tufve Nyholm3, Andrew Leynes4, Peder Larson4, and Florian Wiesinger2

1GE Global Research, Bangalore, India, 2GE Healthcare, Munich, Germany, 3Umeå University, Umeå, Sweden, 4UCSF, San Fransisco, CA, United States

In this work, we demonstrate a generic deep learning (DL) model that computes pCT images (i.e. continuous density bone) using a single channel ZTE MRI data and is robust to protocol and coil variations (as dictated by application needs). The method was evaluated for PET/MR attenuation correction protocol (low resolution for speed) and MRgRTP dose planning protocol (higher resolution for spatial accuracy). The advantages include a single model for multiple protocols, pCT which are very much like real CT in appearance, as well as excellent quantitative accuracy of estimated bone values in the computed pCT.

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

Join Here