ComCat: Combating Covariate Effects in Brain Age Prediction
Poster No:
1604
Submission Type:
Abstract Submission
Authors:
Christian Gaser1, Habib Ganjgahi2, Robert Dahnke1, Thomas Nichols2
Institutions:
1University of Jena, Jena, THU, 2University of Oxford, Oxford, Oxfordshire
First Author:
Co-Author(s):
Introduction:
Accurate estimation of brain age from structural MRI data is often challenged in multi-site studies due to scanner-induced variability, which reduces prediction accuracy when using data from unseen scanners (Gaser et al., 2024). While post-processing methods such as age bias correction can partially address these site effects, harmonization methods such as ComBat are more effective at removing scanner-induced variability in multi-site data (Fortin et al., 2018; Lombardi et al., 2020; Pomponio et al., 2020).
However, the standard ComBat method (Johnson et al., 2007) is limited to correcting for discrete site effects and cannot account for continuous confounding parameters, such as image quality measures (IQMs). To address this limitation, we extended the ComBat approach to develop ComCat (Combating Covariate Effects). ComCat harmonizes imaging data by simultaneously preserving biologically relevant covariates (e.g., age) and removing the effects of both location and IQMs. In this study, we compared ComCat to ComBat and evaluated their impact on brain age prediction accuracy.
However, the standard ComBat method (Johnson et al., 2007) is limited to correcting for discrete site effects and cannot account for continuous confounding parameters, such as image quality measures (IQMs). To address this limitation, we extended the ComBat approach to develop ComCat (Combating Covariate Effects). ComCat harmonizes imaging data by simultaneously preserving biologically relevant covariates (e.g., age) and removing the effects of both location and IQMs. In this study, we compared ComCat to ComBat and evaluated their impact on brain age prediction accuracy.
Methods:
The ComBat harmonization approach is extended by a linear continuous weighted term to model confounding variables whose effects should be removed (e.g., IQMs, Figure 1).
Preprocessing: Gray matter segmentation maps were processed using CAT12 (Gaser et al., 2024b), affine-registered, resampled at 4mm and 8mm, and smoothed with Gaussian kernels. IQMs were derived during preprocessing.
BrainAGE Estimation: BrainAGE was estimated using Gaussian Process Regression (GPR) with four models (4mm/8mm resampling and smoothing combinations), combined as a weighted average (Kalc et al., 2024). Predictions were bias-corrected, and mean absolute error (MAE) was used for evaluation.
ComBat harmonization: Corrected for site effects while preserving age.
ComCat harmonization: Extended ComBat by incorporating IQMs as nuisance variables: 1) Inhomogeneity-to-contrast ratio (ICR); 2) Noise-to-contrast ratio (NCR); 3) Normalized gradient slope of the white matter boundary (res_ECR); 4) RMS error of voxel size (res_RMS); 5) Contrast between tissue classes (contrastr); 6) Overall image quality rating (IQR); 7) structural IQR (SIQR, including res_ECR).
Data: We evaluated ComCat using data from three studies: 1) ABIDE: 437 scans across 14 scanners (ages 10–18); 2) ON-Harmony: 10 participants scanned 80 times on 6 scanners (ages 24–48); 3) Buchert: 1 participant scanned 531 times over 2 years across 116 scanners (age 50).
Preprocessing: Gray matter segmentation maps were processed using CAT12 (Gaser et al., 2024b), affine-registered, resampled at 4mm and 8mm, and smoothed with Gaussian kernels. IQMs were derived during preprocessing.
BrainAGE Estimation: BrainAGE was estimated using Gaussian Process Regression (GPR) with four models (4mm/8mm resampling and smoothing combinations), combined as a weighted average (Kalc et al., 2024). Predictions were bias-corrected, and mean absolute error (MAE) was used for evaluation.
ComBat harmonization: Corrected for site effects while preserving age.
ComCat harmonization: Extended ComBat by incorporating IQMs as nuisance variables: 1) Inhomogeneity-to-contrast ratio (ICR); 2) Noise-to-contrast ratio (NCR); 3) Normalized gradient slope of the white matter boundary (res_ECR); 4) RMS error of voxel size (res_RMS); 5) Contrast between tissue classes (contrastr); 6) Overall image quality rating (IQR); 7) structural IQR (SIQR, including res_ECR).
Data: We evaluated ComCat using data from three studies: 1) ABIDE: 437 scans across 14 scanners (ages 10–18); 2) ON-Harmony: 10 participants scanned 80 times on 6 scanners (ages 24–48); 3) Buchert: 1 participant scanned 531 times over 2 years across 116 scanners (age 50).
Results:
ComCat consistently achieved lower MAE than ComBat in all three studies, even when sites were not explicitly modeled (Figure 2). It removes scanner-induced variance while simultaneously protecting the variables of interest.
Conclusions:
ComCat demonstrated improved harmonization performance. Notably, in the two more realistic scenarios, ComCat outperformed ComBat even without explicitly removing site effects, highlighting its robustness and suitability as a superior harmonization method for predicting brain age.
Lifespan Development:
Lifespan Development Other
Modeling and Analysis Methods:
Multivariate Approaches 1
Novel Imaging Acquisition Methods:
Anatomical MRI 2
Keywords:
Aging
Data analysis
Machine Learning
Modeling
Morphometrics
MRI
Multivariate
Statistical Methods
STRUCTURAL MRI
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
2024;4(10):744-751.
Gaser C, Kalc P, Cole JH. A perspective on brain-age estimation and its clinical promise (2024a). Nat Comput Sci.;4(10):744-751.
Gaser, C., Dahnke, R., Thompson, P. M., Kurth, F., Luders, E., & The Alzheimer's Disease Neuroimaging Initiative (2024b). CAT: a computational anatomy toolbox for the analysis of structural MRI data. GigaScience, 13, giae049.
Johnson WE, Li C, Rabinovic A (2007). Adjusting batch effects in microarray expression data using empirical bayes methods. Biostatistics;8(1):118–127.
Kalc P, Dahnke R, Hoffstaedter F, Gaser C, & Alzheimer's Disease Neuroimaging Initiative (2024). BrainAGE: Revisited and reframed machine learning workflow. Human brain mapping, 45(3), e26632.
Lombardi A, et al. (2020) Extensive Evaluation of Morphological Statistical Harmonization for Brain Age Prediction. Brain Sci.;10(6):364.
Pomponio R, et al. (2020) Harmonization of large MRI datasets for the analysis of brain imaging patterns throughout the lifespan. Neuroimage;208:116450.