Related Experiment Video
Updated: Jan 16, 2026
Inverse Probability of Treatment Weighting Propensity Score using the Military Health System Data Repository and National Death Index
Published on: January 8, 2020
Using Propensity Score Matching to Control for MRI Scan Quality
Veronica J Cramm1, Tyler M Call1, John A E Anderson1
1Department of Cognitive Science, Carleton University.
Abstract:
Movement during MRI scanning complicates distinguishing between the different tissues in the brain (e.g., grey and white matter). Standard practice excludes scans based on researcher-determined visual quality thresholds. Unfortunately, children, elderly, and clinical populations are shown to move more, resulting in higher exclusion rates. This disproportionate exclusion creates systematic bias in the literature and makes research findings less generalizable. Furthermore, the artifacts caused by motion are demonstrated to continue to confound data, even after visual quality control has occurred. We aimed to minimize the confounding factor of systematic group differences in movement. To achieve this, we used a post-scanning statistical technique called propensity score matching (PSM) that matches control and patient populations on scan quality metrics, leading to more comparable groups, greater inclusion, and more generalizable results. We found that PSM can attenuate significant differences in scan quality between groups while allowing for greater sample diversity than standard exclusion protocols. Crucially, using PSM can also alter the results of neuroimaging-based analyses. Using three datasets (total n = 1536), we compared voxel based morphometry analyses based on different quality control protocols. In particular, we observed discrepant results between PSM and strict threshold exclusion, with PSM magnifying some regional group differences and diminishing others. Overall, PSM is a customizable way to mitigate the impact of confounds in neuroimaging research and a powerful method to help distinguish true effects from artifacts.
Insights
Propensity score matching (PSM) reduces bias in neuroimaging by matching MRI scan quality between groups. This method increases sample diversity and can alter analysis results, improving the distinction between true effects and motion artifacts.
Area of Science:
- Neuroimaging
- Biostatistics
- Medical image analysis
Background:
- Motion during MRI scanning complicates brain tissue differentiation and leads to exclusion of scans.
- Standard exclusion protocols disproportionately affect vulnerable populations, introducing systematic bias and reducing generalizability.
- Motion artifacts can persist even after visual quality control, confounding neuroimaging data.
Purpose of the Study:
- To minimize confounding factors from systematic group differences in movement during MRI.
- To evaluate the effectiveness of propensity score matching (PSM) in improving comparability and inclusion in neuroimaging studies.
- To assess the impact of PSM on neuroimaging analysis results compared to standard exclusion methods.
Main Methods:
- Utilized propensity score matching (PSM), a post-scanning statistical technique, to match control and patient populations based on MRI scan quality metrics.
- Compared voxel-based morphometry analyses across three datasets (n=1536) using PSM versus strict threshold exclusion protocols.
- Assessed the ability of PSM to attenuate group differences in scan quality and increase sample diversity.
Main Results:
- PSM successfully attenuated significant differences in scan quality between groups, allowing for greater sample diversity than standard exclusion.
- Neuroimaging analyses using PSM yielded discrepant results compared to strict threshold exclusion, magnifying some regional group differences while diminishing others.
- PSM demonstrated a customizable approach to mitigate confounds in neuroimaging research.
Conclusions:
- Propensity score matching (PSM) is an effective statistical technique for improving group comparability in MRI studies.
- PSM enhances sample diversity and can significantly influence neuroimaging analysis outcomes.
- PSM offers a powerful method to differentiate true biological effects from motion-induced artifacts in neuroimaging research.
Related Concept Videos
01:24Magnetic Resonance Imaging
01:27Imaging Studies IV: Magnetic Resonance Imaging