In NMR in biomedicine ; h5-index 41.0
Increasing the accuracy and reproducibility of subcortical brain segmentation is advantageous in various related clinical applications. In this study, we derived a segmentation method based on a convolutional neural network, U-Net, and a large-scale database consisting of 7039 brain T1-weighted MRI data samples. We evaluated the method by using experiments focused on three distinct topics, namely (1) the necessity of preprocessing steps, (2) cross-institutional and longitudinal reproducibility, and (3) volumetric accuracy. The optimized model MX_RW did not require time-consuming preprocessing steps, such as nonuniform intensity normalization or image registration, for brain MRI before segmentation. Cross-institutional testing revealed that MX_RW (Dice similarity coefficient: 0.809, coefficient of variation: 4.6%, and Pearson's correlation coefficient: 0.979) exhibited performance comparable to that of FreeSurfer (Dice similarity coefficient: 0.798, coefficient of variation: 5.6%, and Pearson's correlation coefficient: 0.973, respectively). The computation time per dataset of MX_RW was generally less than 5 seconds (even when tested without GPUs), which was notably faster than FreeSurfer, which required hours. Thus, for time-restricted applications, MX_RW represents a competitive alternative to FreeSurfer.
Weng Jenn-Shiuan, Huang Teng-Yi
2022-Nov-23
Deep learning, MRI segmentation, subcortical brain structures
