Related Experiment Video
Updated: Jan 12, 2026
17:06
Co-analysis of Brain Structure and Function using fMRI and Diffusion-weighted Imaging
Published on: November 8, 2012
26.9K
Choice of Processing Pipelines for T1-Weighted Brain MRI Impacts Association and Prediction Analyses
Elise Delzant1, Olivier Colliot1, Baptiste Couvy-Duchesne1,2
1Sorbonne Université, Institut du Cerveau - Paris Brain Institute, CNRS, Inria, Inserm, AP-HP, Hôpital de la Pitié-Salpêtrière, Paris, France.
Human Brain Mapping
|October 30, 2025
Summary
Choosing the right MRI processing pipeline is crucial for reliable brain imaging research. Volume-based methods like FSLVBM generally offer better morphometricity, replicability, and prediction accuracy across large datasets.
Area of Science:
- Neuroimaging
- Brain Imaging Research
- Computational Neuroscience
Background:
- Large neuroimaging datasets like the UK Biobank offer opportunities to enhance brain imaging research robustness.
- The influence of different magnetic resonance imaging (MRI) processing pipelines on research outcomes remains largely unknown.
Purpose of the Study:
- To systematically compare five common gray-matter representations from three major software packages (FSL, CAT12/SPM, FreeSurfer).
- To assess the impact of these pipelines on morphometricity, confounder susceptibility, false positives, association findings, and prediction accuracy for 29 diverse traits.
Main Methods:
- Analysis of 39,655 T1-weighted MRI scans from the UK Biobank.
- Comparison of volume-based (FSL) and surface-based (CAT12/SPM, FreeSurfer) gray-matter representations.
- Evaluation across morphometricity, confounder sensitivity, statistical findings, and predictive performance.
Main Results:
- All pipelines showed sensitivity to imaging confounders (e.g., motion, SNR).
- Volume-based methods generally outperformed surface-based methods in detecting significant clusters, replication rates, and predictive performance.
- FSLVBM demonstrated the best overall performance in morphometricity, replicability, and predictive accuracy.
Conclusions:
- Pipeline choice significantly impacts neuroimaging results; FSLVBM is recommended as a consistent all-rounder.
- Caution is advised for interpreting small clusters (single voxels/vertices) due to lower reliability.
- Combining pipelines may enhance prediction, and rigorous confounder management is vital for large-scale neuroimaging studies.
