An Improved SMWI Processing of Substantia Nigra Using Accurate Phase Combination and Deep Neural Network Based QSM

Abstract

Visibility of nigrosome 1, a subregion of substantia nigra is used as an MR imaging biomarker of Parkinsons disease. In this work, we introduced two algorithms for SMWI imaging of substantia nigra. First, we suggested Multi-Channel Phase Combination using Multi-Echo (MCPC-ME), a strategy to calculate and correct phase offsets in multi-echo GRE data. MCPC-ME provided a more accurate estimation of voxel-wise phase offsets particularly in low SNR regions by utilizing phase information from all echoes. Second, we applied QSMnet, a deep neural network for QSM reconstruction, to produce QSM image used in SMWI processing. QSM of nigrosome 1 was reconstructed to have comparable SMWI contrast with 5.4 times faster reconstruction speed compared to the conventional QSM reconstruction algorithm.