Related Experiment Video
Updated: Nov 20, 2025
06:51
PIPEMAT-RS: Development and Validation of a Standardized MATLAB Pipeline for Resting-State EEG Preprocessing
Published on: June 6, 2025
610
Understanding the impact of preprocessing pipelines on neuroimaging cortical surface analyses
Nikhil Bhagwat1, Amadou Barry2, Erin W Dickie3
1Montreal Neurological Institute & Hospital, McGill University, Neurology and Neurosurgery, 3801 University Street, Montreal, H3A 2B4H3A 2B4, Montreal, QC, Canada.
Gigascience
|January 22, 2021
Summary
Neuroimaging preprocessing pipelines significantly impact study results, affecting reproducibility. Understanding these pipeline choices is crucial for accurate analysis of brain structure and function.
Area of Science:
- Neuroimaging
- Computational Neuroscience
- Brain Imaging Analysis
Background:
- Preprocessing pipelines introduce variability in neuroimaging analyses, impacting scientific reproducibility.
- Structural and functional MRI data features are sensitive to algorithmic and parametric differences in preprocessing tasks.
- Mitigating cumulative pipeline biases is critical to distinguish biological effects from methodological variance.
Purpose of the Study:
- To investigate the impact of pipeline selection on cortical thickness measures using open structural MRI datasets.
- To evaluate the effects of different software tools, cortical parcellations, and quality control procedures.
- To analyze how preprocessing choices influence neurobiological group differences and individual prediction tasks.
Main Methods:
- Utilized open structural MRI datasets (ABIDE, Human Connectome Project).
- Investigated effects of software (ANTS, CIVET, FreeSurfer), parcellation (DKT, Destrieux, Glasser), and QC (manual, automatic).
- Performed statistical analyses based on method type (task-free vs. task-driven) and inference objective (group differences vs. individual prediction).
Main Results:
- Software, parcellation, and quality control significantly affect task-driven neurobiological inference.
- Software selection strongly influences both neurobiological (group) and individual task-free analyses.
- Quality control procedures alter performance in individual-centric prediction tasks.
Conclusions:
- Comparative evaluation explains inconsistencies in neuroimaging findings.
- Highlights the need for rigorous scientific workflows and accessible informatics resources.
- Addresses the compounding problem of reproducibility in large-scale computational neuroscience.
