Quality control strategies for brain MRI segmentation and parcellation: Practical approaches and recommendations - insights from the Maastricht study

• DOI: 10.1016/j.neuroimage.2021.118174
• Publication Date: 2021-05-15T11:45:45Z
• Authors: Jennifer Monereo-...; Joost J.A. de Jong; Gerhard S. Drenthen; Magdalena Beran; Walter H. Backes; Coen D.A. Stehouwer; Miranda T. Schram; David E.J. Linden; Jacobus F.A. Jansen
• Keywords: Gold standard (test); Quality Score

Quality control of brain segmentation is a fundamental step to ensure data quality. Manual quality strategies are the current gold standard, although these may be unfeasible for large neuroimaging samples. Several options automated have been proposed, providing potential time efficient and reproducible alternatives. However, those never compared side side, which prevents consensus in appropriate strategy use. This study aimed elucidate changes manual editing segmentations produce morphological estimates, analyze compare effects different on reduction measurement error. Structural MRI from 259 participants The Maastricht Study were used. Morphological estimates automatically extracted using FreeSurfer 6.0. Segmentations with inaccuracies manually edited, before after editing. In parallel, 12 applied full sample. Those included: two strategies, images visually inspected either excluded or edited; five where outliers based tools "MRIQC" "Qoala-T", metrics "morphological global measures", "Euler numbers" "Contrast-to-Noise ratio"; semi-automated detected through mentioned not excluded, but edited. order quantify each strategy, proportion unexplained variance relative total was application resulting differences compared. Manually surfaces produced particularly subcortical volumes moderate cortical surface area, thickness hippocampal volumes. performance depended measure interest. Overall, yielded largest variance. best performing alternatives Euler numbers MRIQC scores. exclusion measures an increase most reliable solution parcellation. must taken prevent subjectivity associated strategies. detection inaccurate provides alternative. avoided.