Abstract #3847

Deep Learning-Based Liver Fat and R2* Mapping with Uncertainty Estimation using Self-Gated Free-Breathing Stack-of-Radial MRI

Shu-Fu Shih^1,2, Sevgi Gokce Kafali^1,2, Tess Armstrong¹, Xiaodong Zhong³, Kara L. Calkins⁴, and Holden H. Wu^1,2

¹Radiological Sciences, University of California, Los Angeles, Los Angeles, CA, United States, ²Bioengineering, University of California, Los Angeles, Los Angeles, CA, United States, ³Siemens Medical Solutions USA, Inc., Los Angeles, CA, United States, ⁴Pediatrics, University of California, Los Angeles, Los Angeles, CA, United States

Self-gated free-breathing multi-echo stack-of-radial MRI quantifies liver fat and R₂*. However, data undersampling due to motion self-gating can degrade the image quality and quantification accuracy. Previous methods required either longer scan time or computationally expensive constrained reconstruction. In this work, a deep learning-based two-stage network was developed to suppress undersampling artifacts and rapidly generate quantitative fat and R₂* maps with a pixel-wise uncertainty map. The proposed method achieved accurate fat and R₂* mapping and reduced the computational time by two orders of magnitude versus constrained reconstruction. The uncertainty map can be used to detect regions with potential quantification errors.

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

Join Here