Imputation-Based Harmonization Mitigates Site Effects Without Data Leakage in Machine Learning Studies

Topic: Brain imaging, connectivity and assessment

Neuroimaging studies that pool data across clinical sites often suffer from site effects — variability in imaging measures that arises from technical heterogeneity across sites as opposed to true biological signal. While numerous harmonization methods have been proposed to remove site effects from imaging features, less attention has been placed on how to incorporate harmonization models into common machine-learning pipelines. We demonstrate that current approaches for integration either suffer from data leakage, fail to fully remove site effects from the target data, or attenuate true biological associations during the harmonization process. To address these issues, we propose Multiple Imputation for Removing Technical Heterogeneity (MIRTH), which uses imputed outcomes to harmonize the test data. We benchmark MIRTH’s performance using both simulated and real-world volumetric data from the Alzheimer’s Disease Neuroimaging Initiative and the Baltimore Longitudinal Study of Aging, demonstrating that MIRTH can achieve high predictive accuracy while avoiding inflated performance when the outcome is imbalanced across clinical sites.

Key findings

• Neuroimaging studies that pool data across clinical sites often suffer from site effects — variability in imaging measures that arises from technical heterogeneity across sites as opposed to true biological signal.
• While numerous harmonization methods have been proposed to remove site effects from imaging features, less attention has been placed on how to incorporate harmonization models into common machine-learning pipelines.
• We demonstrate that current approaches for integration either suffer from data leakage, fail to fully remove site effects from the target data, or attenuate true biological associations during the harmonization process.
• To address these issues, we propose Multiple Imputation for Removing Technical Heterogeneity (MIRTH), which uses imputed outcomes to harmonize the test data.
• We benchmark MIRTH’s performance using both simulated and real-world volumetric data from the Alzheimer’s Disease Neuroimaging Initiative and the Baltimore Longitudinal Study of Aging, demonstrating that MIRTH can achieve high predictive accuracy while avoiding inflated performance when the outcome is imbalanced across clinical sites.

New to this subject? Start with our plain-English guide: Digital health →

Study details

• Study designlongitudinal study
• Evidence strengthModerate

Read the original

• DOI: 10.64898/2026.09.28.755204
• Original source →

Free copy unknown: no free-copy check has an answer for this record yet. That does not mean it is paywalled.

@article{hillman2026-35cc200a,
  author  = {Hillman, N. and Chen, A. A. and Hu, F. and Vandekar, S. N. and Melhem, R. and Beason-Held, L. L. and Satterthwaite, T. D. and Davatzikos, C. and others},
  title   = {{Imputation-Based Harmonization Mitigates Site Effects Without Data Leakage in Machine Learning Studies}},
  journal = {bioRxiv},
  year    = {2026},
  doi     = {10.64898/2026.09.28.755204},
  url     = {https://doi.org/10.64898/2026.09.28.755204}
}

Related studies in Digital health tools

• Mitigating site effects in covariance for machine learning in neuroimaging data Human brain mapping · 2021-12-14 · usefulness score 43/100
• IRMA: Machine learning-based harmonization of $$^{18}$$F-FDG PET brain scans in multi-center studies European Journal of Nuclear Medicine and Molecular Imaging · 2025-02-18 · usefulness score 41/100
• Explainable machine learning for predicting longitudinal dementia status: Establishing a leakage-free benchmark PLOS Digital Health · 2026-05-18 · usefulness score 50/100
• Bridging the Gap: Missing Data Imputation Methods and Their Effect on Dementia Classification Performance Brain Sciences · 2025-06-13 · usefulness score 40/100
• Machine Learning-Based Alzheimer’s Disease Stage Diagnosis Utilizing Blood Gene Expression and Clinical Data: A Comparative Investigation Diagnostics · 2025-01-17 · usefulness score 36/100
• Enhanced Multi-Model Machine Learning-Based Dementia Detection Using a Data Enrichment Framework: Leveraging the Blessing of Dimensionality Bioengineering · 2025-05-30 · usefulness score 34/100

Similar studies

• Diagnostic Labels and Measurement Timing Drive Systematic Inconsistency in ADNI Neuroimaging Data 2026 · Digital health tools
• Neuro-JEPA: Building and Evaluating Health-System Scale Neuroimaging Foundation Models with Mixture of Experts 2026 · Mild cognitive impairment
• Comparing Harmonization Approaches for Protocol-Related Variability in Multisite Diffusion MRI Data 2026 · Brain scans and imaging
• Longitudinal ComBat: A method for harmonizing longitudinal multi-scanner imaging data 2020 · Brain scans and imaging
• A Pipeline for ADNI Resting-State Functional MRI Processing and Quality Control 2026 · Long-term cohort studies
• Current gaps in downstream applications of fMRI foundation models: lessons learned from BrainLM & BrainHarmonix on Prevent-AD 2026 · Brain scans and imaging

More on Digital health tools

• A review that covers this ground Machine learning for modeling the progression of Alzheimer disease dementia using clinical data: a systematic literature review JAMIA open · 2021-07-01 · usefulness score 44/100
• Newest in this subject Speaker-Wise Dementia Screening fromNaturalistic Speech with Lexical Entropy, Semantic Drift, and Multimodal Fusion Research Square · 2026-10-21 · Preprint — not yet peer-reviewed · usefulness score 16/100
• Newest in this subject Integrating Multisource Target Evidence with Machine-Learning Potency Prediction for Alzheimer’s Disease Drug Discovery bioRxiv · 2026-10-20 · Preprint — not yet peer-reviewed · usefulness score 16/100

See all studies in Digital health tools →