Eliminate batch effects.
Preserve biological signal.
Robust longitudinal harmonization for multi-center brain MRI studies.
longCombat bridges the gap between imaging centers.
A specialized toolbox that corrects scanner- and site-specific variations in brain MRI while preserving true biological and longitudinal changes in studies spanning multiple hospitals and timepoints.
Multi-center MRI data is not directly comparable.
Differences in scanner vendors, field strength, acquisition protocols, and reconstruction algorithms introduce substantial non-biological variance. In longitudinal studies this variance can mask or mimic disease progression signals, particularly in studies across various neurological conditions.
Standard batch correction methods often fail when applied to repeated measures or when the number of subjects per site is limited — common realities in clinical neuroimaging research.
longCombat combines ComBat with longitudinal modeling.
We extend the ComBat framework with linear mixed-effects modeling to explicitly account for within-subject correlation across time while removing site effects.
Models both cross-sectional site effects and subject-specific trajectories over time using linear mixed-effects, preventing over-correction of true longitudinal signals.
Automatically creates study-specific consensus brain masks that maximize tissue coverage while minimizing site-specific artifacts across all participating centers.
Built for real-world multi-center studies
Explicitly models within-subject correlation across multiple timepoints using LME, preserving disease progression signals while removing scanner effects.
Extended ComBat implementation supporting arbitrary numbers of sites with robust empirical Bayes shrinkage even when some centers have small sample sizes.
Generates optimal brain masks from all sites combined, ensuring consistent tissue segmentation and reducing partial volume effects in downstream VBM and DTI analyses.
Full-featured graphical interface for non-programmers alongside powerful command-line and scriptable MATLAB/SPM.
Extensively tested to eliminate variations introduced by different scanners, centers, and hospitals. Suitable for a wide range of diseases, with quantitative comparison against standard ComBat and other harmonization methods.
Generates harmonized NIfTI images, quality control reports, voxel-wise statistics, and complete provenance logs for publication-ready results.
From raw multi-center data to harmonized images in minutes.
Load longitudinal NIfTI data from multiple centers. Automatic detection of timepoints and site labels.
Create study-specific brain mask optimized across all sites and timepoints for consistent analysis space.
Apply longitudinal ComBat + LME correction. Adjust for age, sex, disease severity, and other covariates.
Review voxel-wise statistics, TFCE results, and export harmonized images ready for VBM, DTI, or statistical modeling.
Proven impact on real multi-center studies.
Ready to harmonize your data?
Complete user guide, parameter reference, and step-by-step tutorials for the MATLAB interface.
Read the docsFull source code, example datasets, validation scripts, and issue tracker for community contributions.
View on GitHubPublicly available multi-center longitudinal MRI datasets pre-processed with longCombat.
Download examples