Meeting Banner
Abstract #3194

Feasibility of magnetic resonance based electrical properties tomography with deep learned reconstruction based denoising

Ho-Joon Lee1, Yeonah Kang1, Marc Lebel2, Joon-Hyeong Kim3, Dong-Hyun Kim3, and Sung-Min Gho4
1Department of Radiology, Haeundae Paik Hospital, Busan, Republic of Korea, 2MR Collaboration and Development, GE Healthcare, Calagary, AB, Canada, 3Department of Electrical and Electronic Engineering, Yonsei University, Seoul, Republic of Korea, 4MR Collaboration and Development, GE Healthcare, Seoul, Republic of Korea

With advances in deep learning, feasibility has been investigated for MREPT reconstruction showing interesting results. However whether images denoised with deep learned reconstruction will improve EPT map quality has not been investigated. After denoising of complex data acquired with a DL algorithm, EPT maps were generated with phase based 2D-weighted polynomial fitting. Use of DL, shows better results as compared to conventionally generated maps (i.e. decreased NRMSE, increased PSNR and SSIM, with increasing denoising levels), and results in sharper appearing maps. Spreading of boundary artifacts are not observed with increasing denoising factors.

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

Join Here