Harmonization Remains Unresolved: The Case of White Matter Hyperintensity Using ComBat And Deming Regression

Topic: Brain proteins, microglia and lab models

Abstract Background Multi‐site aging and dementia studies are challenged by data heterogeneity caused by scanner (manufacturer and protocol) related variability. Quantitative analysis of white matter hyperintensity (WMH) from pooled samples, a traditional marker of cerebrovascular disease, is sensitive to machine variability. Our goal was to compare Deming (symmetric) linear regression and ComBat, a common method used for harmonizing neuroimaging measurements, for harmonizing WMH measurements on participants scanned on two different scanners (GE and Siemens) and protocols (2D FLAIR on GE and 3D FLAIR on Siemens). Methods Cognitively unimpaired (CU) ( n =53, age (mean (SD))=67 (15) years) and cognitively impaired (CI; MCI/dementia) ( n =34, age (mean (SD))=82 (9) years) participants from the Mayo Clinic Study of Aging and Mayo ADRC were scanned on GE and Siemens scanners within a week of each other (typically done on the same day). WMH as a percentage of total intracranial volume was measured using multispectral segmentation of FLAIR and T1‐weighted images (Gunter et. al. AAIC 2023). We used Deming regression as it includes errors in both input dimensions. ComBat was applied using age, sex, and diagnosis as covariates, with scanner vendor and protocol as the batch variable. We assessed the inter‐scanner similarity using Maximum Mean Discrepancy (MMD) analysis, where a value of zero indicates identical distributions, p >0.05 indicates no significant difference under H 0 . (Gretton et al., JMLR 2012). Results Our results showed that the distribution of WMH measurements, which were different in the CU group, were more similar after Deming than ComBat harmonization (Figure 1). This was confirmed by the lower MMD in the CU group from Deming regression (‐0.01 ( p =0.76) from 0.03 ( p =0.05)) compared to ComBat (0.04 ( p =0.02))(Figure 1). However, ComBat and Deming were unable to create distributions as compatible as original measurements for the CI group. Conclusion Using a unique dataset of the same participants scanned close in time on different scanner and protocols, we found that (1) MMD measurements were useful to test the need/improvement due to harmonization; (2) ComBat was not substantially better at creating compatible distributions between scanner vendors than Deming regression, and both were sometimes inadequate. Future work will require alternative models to linear scale and shift transformations.

Key findings

• We assessed the inter‐scanner similarity using Maximum Mean Discrepancy (MMD) analysis, where a value of zero indicates identical distributions, p >0.05 indicates no significant difference under H 0 .
• Results Our results showed that the distribution of WMH measurements, which were different in the CU group, were more similar after Deming than ComBat harmonization (Figure 1).
• This was confirmed by the lower MMD in the CU group from Deming regression (‐0.01 ( p =0.76) from 0.03 ( p =0.05)) compared to ComBat (0.04 ( p =0.02))(Figure 1).
• Conclusion Using a unique dataset of the same participants scanned close in time on different scanner and protocols, we found that (1) MMD measurements were useful to test the need/improvement due to harmonization; (2) ComBat was not substantially better at creating compatible distributions between scanner vendors than Deming regression, and both were sometimes inadequate.

Why it matters

This methods study compared two ways of harmonizing white matter hyperintensity measurements taken on different scanners and found that Deming regression made the distributions more similar than ComBat in cognitively unimpaired participants. It addresses how to compare imaging data across sites in dementia research, but it does not test any treatment, cure, or prevention.

Study details

• Participants53
• Evidence strengthLow

Read the original

• DOI: 10.1002/alz70861_108630
• PubMed: 41434809
• Original source →
• Free full text: Alzheimer s & Dementia →

Full text: free to read — OpenAlex lists this copy as free to read: the link above (checked 6 Feb 1020).

@article{reyes2025-41434809,
  author  = {Reyes, Calvin D. and Gebre, Robel K and Gunter, Jeffrey L. and Hu, Mingzhao and Schwarz, Christopher G. and Meyer, Nolan K. and Reid, Robert I. and Kantarci, Kejal and others},
  title   = {{Harmonization Remains Unresolved: The Case of White Matter Hyperintensity Using ComBat And Deming Regression}},
  journal = {Alzheimer s \& Dementia},
  year    = {2025},
  doi     = {10.1002/alz70861_108630},
  pmid    = {41434809},
  url     = {https://doi.org/10.1002/alz70861_108630}
}

Related studies in Vascular contributions

• Attention Blocks Improve White Matter Hyperintensity Semantic Segmentation using U-Nets IEEE Latin America Transactions · 2025-07-08 · usefulness score 30/100
• Investigating links between white matter hyperintensities and menopausal status using robust age-correction methods in UK Biobank Brain Communications · 2025-01-01 · usefulness score 49/100
• Characterizing spatiotemporal white matter hyperintensity pathophysiology in vivo to disentangle vascular and neurodegenerative contributions Nature Communications · 2026-03-31 · usefulness score 58/100
• Insomnia, early and late rising are associated with small hippocampal volume and large white matter hyperintensity burden Alzheimer s Research & Therapy · 2025-04-05 · usefulness score 57/100
• Additive Effects of White Matter Hyperintensity and APOE ε4 Status on Risk of Incident Dementia in Two Large Longitudinal Cohorts Annals of Neurology · 2025-11-21 · usefulness score 57/100
• Associations of Handgrip Strength and Testosterone With Cerebral White Matter Hyperintensity and Microstructural Injury Journal of Cachexia Sarcopenia and Muscle · 2025-06-01 · usefulness score 56/100

Also tagged “cerebral”

900 studies in the library carry the topic tag “cerebral”; these are this study's neighbours in a fixed, alphabetical-by-hash tour of them, not the closest in content. All 900 tagged studies.

• Molecular mechanisms of cognitive dysfunction following traumatic brain injury Frontiers in aging neuroscience · 2013-07-09 · usefulness score 40/100 · Disease mechanisms
• A more irregular shape of white matter hyperintensities is associated with cognitive decline over five years in community-dwelling older adults Neurobiology of Aging · 2025-04-03 · usefulness score 46/100 · Vascular contributions
• Testing the Association Between Neighborhood Disadvantage and White Matter Hyperintensities Alzheimer s & Dementia · 2022-12-01 · usefulness score 37/100 · Risk factors

Similar studies

• Association of Data-Driven White Matter Hyperintensity Spatial Signatures With Distinct Cerebral Small Vessel Disease Etiologies 2022 · Biomarkers
• White matter hyperintensity spatial patterns: Risk factors and clinical correlates 2025 · Risk factors
• Coevolution of white matter hyperintensities and cognition in the elderly 2012 · Brain scans and imaging
• Impact of white matter hyperintensity volume progression on rate of cognitive and motor decline 2008 · Caregiving
• Neuropathologic basis of white matter hyperintensity accumulation with advanced age 2013 · Brain scans and imaging
• Regional distribution of white matter hyperintensities in vascular dementia, Alzheimer's disease and healthy aging 2004 · Cognitive tests and assessment

More on Vascular contributions

• A review that covers this ground White matter hyperintensities and risks of cognitive impairment and dementia: A systematic review and meta-analysis of 36 prospective studies Neuroscience and biobehavioral reviews · 2020-11-11 · usefulness score 79/100
• Newest in this subject Precision Nanocrystal Theranostics for Neurodegenerative Proteopathies: Strategies to Cross the Blood-Brain Barrier and Target Protein Aggregation Preprints.org · 2026-10-26 · Preprint — not yet peer-reviewed · usefulness score 16/100
• Newest in this subject Depth-Dependent Alterations in Cerebral Blood and Tissue Oxygenation Associated with Tau in Alzheimer’s Disease Models bioRxiv · 2026-10-16 · Preprint — not yet peer-reviewed · usefulness score 16/100

See all studies in Vascular contributions →