Reproducibility Evaluation of SLANT Whole Brain Segmentation Across Clinical Magnetic Resonance Imaging Protocols

摘要

Whole brain segmentation on structural magnetic resonance imaging (MRI) is essential for understandingneuroanatomical-functional relationships. Traditionally, multi-atlas segmentation has been regarded as the standardmethod for whole brain segmentation. In past few years, deep convolutional neural network (DCNN) segmentationmethods have demonstrated their advantages in both accuracy and computational efficiency. Recently, we proposedthe spatially localized atlas network tiles (SLANT) method, which is able to segment a 3D MRI brain scan into 132anatomical regions. Commonly, DCNN segmentation methods yield inferior performance under external validations,especially when the testing patterns were not presented in the training cohorts. Recently, we obtained a clinicallyacquired, multi-sequence MRI brain cohort with 1480 clinically acquired, de-identified brain MRI scans on 395patients using seven different MRI protocols. Moreover, each subject has at least two scans from different MRIprotocols. Herein, we assess the SLANT method’s intra- and inter-protocol reproducibility. SLANT achieved lessthan 0.05 coefficient of variation (CV) for intra-protocol experiments and less than 0.15 CV for inter-protocolexperiments. The results show that the SLANT method achieved high intra- and inter- protocol reproducibility.