The processing steps involved in MR elastography include custom imaging sequences, image reconstruction, and material property estimation, but the most person-hours are spent on manual image masking. Manual corrections are needed because automated brain segmentation often fails near temporal lobe artifacts, which are unique for each subject. A deep learning method, specifically, a U-net architecture, is trained to map input MRE-image intensity data to corresponding manually corrected masks (N=44) with the goal of automating the masking of future MRE brain datasets. We observed the U-net-based masking maintained data quality (OSS-SNR) for all subjects.
This abstract and the presentation materials are available to members only; a login is required.