Search bioRxiv⌕ Search

bioRxiv · 10.64898/2026.05.26.727928

Superpixel-ComBat multi-site harmonization of unpaired T1W MRI data in Huntingtons disease

Abstract

BackgroundPooling multi-site MRI data is essential for well-powered neuroimaging analyses, particularly in Huntingtons disease (HD), where large cohorts are needed to study disease-stage heterogeneity and subtle progressive neuroanatomical change. However, scanner-related variability hinders direct data pooling, confounding image-level methods such as voxel-based morphometry (VBM). Superpixel-ComBat (SP-ComBat), a voxel-level image-harmonization framework, effectively removes scanner effects but depends on traveling-subject data that are rarely available retrospectively. We extend SP-ComBat to unpaired multi-site T1-weighted MRI by introducing a pseudo-pairing framework that leverages demographically matched controls across scanners as surrogate traveling subjects. MethodsTwo pipelines were developed to estimate scanner effects under retrospective constraints: pipeline 1 used a small set of well-matched pseudo-pairs (n = 4) with bootstrap resampling to address scanners with limited sample sizes, while pipeline 2 used the recommended number of pseudo-pairs (n = 16) without resampling. Pseudo-pair images were parcellated into 3D-superpixels, and ComBat was applied within clusters to estimate scanner-specific adjustments for native-space harmonization. Pipeline performance was assessed in a representative multi-study dataset comprising six scanners from three HD cohorts (HD-YAS, HD-CSF, TRACK-HD; N = 144) and replicated in the full multi-study dataset (FMD; N = 548). ResultsBoth pipelines improved image quality, aligned scanner-specific intensities, and preserved disease-related structural patterns. Pipeline 2 showed superior parameter re-estimation stability and was selected for the FMD. Harmonization eliminated systematic segmentation errors, enabled a single unified VBM pipeline across scanners, and increased sensitivity to HD-related voxel-wise atrophy. ConclusionsSP-ComBat was effectively adapted for harmonization of unpaired multi-site structural MRI, reducing scanner bias while preserving biological variability and supporting unified VBM analyses across scanners.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Coleman, A., Chen, C.-L., Hanson-Baiden, J., Minhas, D. S., Torbati, M. E., Laymon, C. M., Tabrizi, S. J., Wild, E. J., Tudorascu, D. L., Scahill, R. I., Byrne, L. M.. 2026-05-29. Superpixel-ComBat multi-site harmonization of unpaired T1W MRI data in Huntingtons disease. https://doi.org/10.64898/2026.05.26.727928

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

Bullying Victimisation, Medial Occipital White Matter Development, and Psychotic-Like Experiences: A Longitudinal Diffusion MRI Study

Childhood bullying victimisation is robustly associated with psychotic-like experiences, but the neurodevelopmental mechanisms underlying this association remain unclear. This longitudinal study examined whether childhood bullying victimisation was associated with altered maturation of medial occipital white matter and whether bullying victimisation moderated the association between fractional anisotropy (FA) and psychotic-like experiences. A total of 96 young people (38 bullied, 58 controls) from the Adolescent Brain Development study underwent diffusion MRI at approximately 14, 16, and 20 years of age. PLE and bullying victimisation group classification were determined by expert consensus interview. FA from bilateral calcarine-cuneus, cuneus-lingual, and calcarine-lingual connections was examined using mixed-effects models. Compared with controls, the bullied group showed significantly different age-related FA trajectories in the right calcarine-cuneus connection ({beta} = -0.40, p < .001) and the left cuneus-lingual connection ({beta} = -0.35, p = .003). Bullying victimisation significantly moderated the association between white matter FA and PLE ({beta} = -6.14, p = .004), indicating that the association between FA and psychotic-like experiences differed according to bullying victimisation status. This moderation effect also varied with age, as indicated by a significant bullying victimisation-by-FA-by-time interaction ({beta} = -6.80, p = .008). Specifically, the moderating effect of bullying victimisation on the association between right calcarine-cuneus FA and PLE changed across development. These findings suggest that bullying victimisation is associated with divergent maturation of medial occipital circuitry and identify right calcarine-cuneus development as a potential marker of vulnerability to PLE following childhood bullying.

neuroscience↗

The social conditioned place preference assay is sensitive to drug dose, training regimen, and social context

Beginning in the 1960s, researchers have understood that one of the unique properties of psychedelic drugs is that their effects are modified by the context in which they are taken (also known as their set and setting dependence). To date, 7 different studies have shown that critical period reopening is a shared property across psychedelic drugs, including ketamine, MDMA, psilocybin, LSD, and ibogaine. Moreover, the context dependence of psychedelics effects has been recapitulated for both MDMA and ketamine. Recently, Lu et al., Nature Neuroscience 2026 (PMID: 42972049) have claimed that psilocybin does not reopen the social reward learning critical period, which the authors interpret as a failure to replicate previously published results. Here, we re-analyzed results across three published datasets (Nardou et al., Nature 2019; Nardou et al., Nature 2023; Lu et al., Nature Neuroscience 2026) in order to compare the impact of dose, training regimen, and social context on the social conditioned place preference (social CPP) assay. Our analysis reveals that methodological modifications likely account for Lu and colleagues inability to reopen the social reward learning critical period with psilocybin. Nevertheless, because their results recapitulate previous findings that psychedelics effects are context dependent, they support rather than challenge the growing literature demonstrating that psychedelics reopen critical periods.

neuroscience↗

Harmonizing amyloid PET tracers at the local level: evaluation of regional Centiloid equations

INTRODUCTION: The Centiloid (CL) framework standardizes global amyloid-PET across radiotracers. While global scaling remains essential for participant selection, amyloid-targeting therapies require quantifying localized plaque removal. However, applying global equations regionally introduces systematic bias, particularly in subcortical structures where off-target binding alters non-specific background. METHODS: We extended the Centiloid framework regionally and voxel-wise using GAAIN datasets: Level-1 ([11C]PiB) and Level-2 paired [11C]PiB and fluorinated head-to-head cohorts ([18F]florbetaben, [18F]florbetapir, [18F]flutemetamol, and [18F]NAV4694). Level-1 defines the regional Centiloid scale using region-specific intercepts while fixing the global slope. Level-2 calibrates fluorinated tracer uptake to reference [11C]PiB SUVR using linear equations. We compared three slope-intercept models (Global, Regional-Intercept, and Regional) using information criteria and Mean Absolute Error (MAE). RESULTS: Regional-Intercept model was identified as optimal across all four radiotracers. In neocortical regions, Global and Regional-Intercept models showed high agreement across canonical lobes. In deep gray matter, global equations introduced large systematic offsets, with mean absolute conversion errors in the pallidum reaching 27.7 to 45.6 CL. The Regional-Intercept model eliminated this bias, reducing subcortical MAE on average below 15 CL while preserving a unified biological scale. DISCUSSION: Applying global Centiloid equations directly to regions introduces substantial bias in subcortical regions. With the Regional-Intercept model, the same Centiloid value corresponds to the same amyloid concentration across brain regions and radiotracers. This provides a reliable basis for cross- sectional regional evaluation, multi-tracer cohort pooling, tracking regional clearance, and mapping regional progression in clinical research.

neuroscience↗