Optimizing FreeSurfer’s Surface Reconstruction Parameters for Anatomical Feature Estimation

• DOI: 10.1101/2023.01.02.522457
• Publication Date: 2023-01-03T16:36:05Z
• Authors: Zvi Baratz; Yaniv Assaf
• Keywords: Fluid-attenuated inversion recovery; Similarity (geometry); Feature (linguistics); Statistical power; Data pre-processing; Sample (material)

Abstract Magnetic resonance imaging (MRI) is a powerful tool for non-invasive of the human body. However, quality and reliability MRI data can be influenced by various factors, such as hardware software configurations, image acquisition protocols, preprocessing techniques. In recent years, introduction large-scale neuroimaging datasets has taken an increasingly prominent role in neuroscientific research. The advent publicly available standardized repositories enabled researchers to combine from multiple sources explore wide range scientific inquiries. This increase scale allows study phenomena with smaller effect sizes over more diverse sample greater statistical power. Other than variability inherent across sites, feature generation steps implemented different labs introduce additional layer which may influence consecutive procedures. this study, we show that differences configuration surface reconstruction anatomical using FreeSurfer results considerable changes estimated features. addition, demonstrate these have on within-subject similarity performance basic prediction tasks based derived Our although provided either T2w or FLAIR scan same purpose improving pial estimation (relative mandatory T1w alone), two configurations distinctly effect. our findings indicate scans models sex age are significantly effected, they not improved enhanced configurations. These large extent elementary sparsely reported workflow brain meant underline importance optimizing procedures experimental prior their distribution standardization harmonization efforts public datasets. should carefully included any following analytical workflows, account variation originating differences. Finally, other representations raw explored studied provide robust framework aggregation sharing.