Comparison of transform and image domain denoising techniques for multi-slice multi-echo T2 relaxation time mapping

Motivation: The SNR efficiency of multi-slice multi-echo (MSME) T2 mapping can be improved by denoising strategies. Goal(s): To evaluate the performance of a recently proposed low-rank transform domain technique vs. prior-art image domain denoising techniques for MSME T2 mapping. Approach: A reference 20-average T2 mapping data set of a piglet knee specimen was acquired at 3T. The performances of transform (T-MPPCA and T-NORDIC) and image domain (MPPCA and BM4D) denoising were compared at different SNR levels and for magnitude vs. complex image data. Results: T-NORDIC best preserved T2 map spatial fidelity and quantitative accuracy while providing a substantial reduction in image noise. Impact: T-NORDIC provides significant advantages for denoising of MSME T2 mapping acquisitions compared to image domain methods, including improved quantitative accuracy and spatial fidelity, providing a robust approach to improve SNR efficiency while being conservative and preserving subtle local texture.