Biology subjects

Zhou, J. H.

Publications and source records attributed to Zhou, J. H..

At least 19 recordsLinked to original sources

bioRxiv2026

Diet Quality from Midlife to Later Life Relates to Late-Life Brain Health and Verbal Memory in the SG70 Cohort

Healthy diet across adulthood is associated with better late-life cognition, but how life-course diet quality relates to brain integrity, and whether brain measures mediate diet-cognition associations, remains unclear as studies with long-term diet records and detailed neurocognitive measures are lacking. We studied 892 participants from the SG70 study, nested within the Singapore Chinese Health Study, with adherence to the Dietary Approaches to Stop Hypertension diet (DASH) assessed between 1993--2025. Dietary quality during midlife, ages 44--55 years, and early elderhood, ages 61--73 years, was examined in relation to seven cognitive domains, brain morphometry, white matter hyperintensities and free-water MRI markers in late life, ages 68--82 years. Higher DASH adherence at both life stages was significantly associated with better late-ife verbal memory, and remained so when both life stages were modelled jointly. Higher midlife DASH adherence was associated with greater white matter volume in association tracts, whereas higher early-elderhood DASH adherence was associated with lower white matter hyperintensity (deep basal ganglia and anterior periventricular regions) and lower frontal and occipital grey matter free water, suggesting lower neurovascular and inflammatory burden. Mediation analyses indicated that white matter volume accounted for 12.3% in mediating the midlife DASH--verbal memory association, while cortical free water accounted for 12.5% in mediating the early-elderhood DASH--verbal memory association. Importantly, participants whose DASH adherence improved from lower adherence in midlife to better adherence in later life showed better verbal memory and more favourable brain integrity than those with persistently low adherence. These findings identify midlife and post-midlife diet quality as modifiable life-course exposures associated with late-life cognitive resilience through differences in macrostructural and microstructural brain integrity.

neuroscience↗

bioRxiv2026

From Geometry to Hierarchy: Charting the Dominant Modes of Human Cortical Function Through Development

Cortical function in the human cortex is spatially patterned along a macroscale information processing hierarchy. This hierarchy tracks a sensory-association axis anchored at one end by unimodal sensory regions and at the other by transmodal association areas. Biophysical models of brain dynamics suggest that this axis should align with the dominant, resonant modes of cortical anatomy and geometry, but they do not. Instead, the dominant anatomical modes correspond to simple spatial gradients extending along rostro-caudal, medio-lateral, and dorso-ventral axes. How does this divergence between anatomical and functional modes of the cortex arise? Here, using four independent functional magnetic resonance imaging datasets spanning infancy to young adulthood (n=1,512, aged 0-21 years), we show that this spatial divergence is attributable to the maturation of long-range cortico-cortical inter-regional functional coupling. Specifically, we find that the dominant functional mode of the human cortex during the first 3 to 4 years of life closely resembles the rostro-caudal mode of cortical geometry, then shifts to a rudimentary sensory-association-like mode between 4 and 6 years of age and ultimately matures into a prototypical sensory-association mode by age 13. In infancy and early childhood, a linear model predicting the topography of the dominant functional mode from modes of cortical geometry substantially outperforms a model relying on the adult sensory-association mode, whereas the reverse is true in later stages of development. Furthermore, targeted removal of long-range connections causes the adult functional mode to regress from a sensory-association topography to an infant-like rostro-caudal mode. Our findings indicate that the maturation of long-range cortico-cortical coupling supports the gradual emergence of hierarchical modes of cortical function from an organization initially dominated by geometric constraints.

neuroscience↗

bioRxiv2026

Widespread use of invalid statistical tests in biomedical machine learning

Cross-validation is routinely used to compare performance in biomedical artificial intelligence. Standard tests ignore correlation across cross-validation folds, inflating false-positive rates. In a PRISMA-guided meta-analysis of 184 studies (impact factor [≥] 15) across 30 biomedical fields, 97% use invalid tests. Among studies with abstract-level claims supported by invalid tests, 59% rely on a spurious comparison: one that loses significance after correcting for this correlation. On average, regaining significance requires a 43% larger effect. Extending these findings to the broader literature, we estimate that spurious comparisons occur in nearly two in three studies and support abstract-level claims in nearly one in three studies. Simulations confirm false-positive rates paradoxically approach 100% when cross-validation is repeated to improve stability. We introduce SHARP, a redesign of cross-validation, which best balances false-positive control and power among 13 benchmarked tests. These results reveal widespread fragility in biomedical artificial intelligence and offer a practical route to valid comparisons.

bioinformatics↗

bioRxiv2026

Developing a multi-modal neuroimaging-based BrainAge model across childhood

BrainAge models hold promise as a clinical biomarker for developmental brain health, especially in childhood when there is the potential for early intervention. To distinguish between normative developmental variance and pathological divergence, BrainAge models should reflect the dynamic and diverse neurodevelopmental processes that occur in distinct developmental windows across childhood. We utilized multi-modal neuroimaging data from three pediatric cohorts covering ages 4 to 13 years (n = 1005, 2126 scans), split into Train and Test datasets. Twelve sex-stratified BrainAge models were built stratified by type and different combinations of neuroimaging features. Model types were "Full-Span" models covering the full age range, and "Phase-Specific" models split into early- and late-childhood. We first compared BrainAge estimates in the Test dataset amongst our candidate models, then benchmarked the best-performing model against published pre-trained models and DNA-based biological age measures. Our findings show that a BrainAge model that was phase-specific and consisted of both structural and functional features (cortical thickness, subcortical volumes, and functional network integration measures) showed good prediction of age and best distinguished between healthy and symptomatic subgroups. We present a proof-of-concept for developmental models supporting building BrainAge models of higher temporal resolution that align to different childhood developmental phases.

neuroscience↗

bioRxiv2026

Sequential narrative binding by hippocampal CA2/3 sustains lifespan episodic detail retrieval

Hippocampal subfield contributions to the structure of episodic memories across the lifespan remain incompletely understood, particularly regarding the mechanisms underlying sustained retrieval of detailed events regardless of age. Here, we show that focal hippocampal damage in 32 individuals with amnesia produces time-invariant loss of episodic content details (event specifics, place, perceptual, and emotional content) across six decades, sparing only early childhood memories and semantic details. Temporal context details showed non-monotonic vulnerability, with maximal deficits in recent memories. Volumetric analyses revealed distinct anterior-posterior functional gradients: anterior hippocampal CA3 predicted retrieval across 11-55+ years, posterior CA3 across 18-55+ years, whereas CA1 predicted only recent memories. Computational linguistic analyses using Sentence-BERT semantic embeddings quantified narrative organisation within remembered events via consecutive-sentence semantic relatedness (local narrative coherence) and overall semantic alignment (global narrative coherence). Hippocampal CA2/3 damage impaired local but not global narrative coherence, with reduced local coherence mediating episodic detail loss. This dissociation demonstrates that CA2/3 mediates sequential narrative binding through mechanisms dissociable from broader narrative operations. Concordant time-invariant deficits across quantitative measures of episodic content and computational-linguistic based measures challenge classical systems consolidation predictions, supporting parallel hippocampal-cortical contributions to lifespan episodic memory.

neuroscience↗

bioRxiv2025

Decoding memory dysfunction in hippocampal amnesia

The hippocampus reconstructs past experiences by integrating sensory, perceptual, and conceptual information across a cortico-hippocampal autobiographical memory network (AMN). Here, in 18 human participants with amnesia, we decoded the effects of bilateral focal hippocampal damage on distinct autobiographical representations using representational dissimilarity matrices (RDMs). Hippocampal pathology resulted in: (1) impaired generalised episodic memory retrieval RDM model fit in the left angular gyrus; and, (2) reduced distinct episodic memory RDM model fit and trial-by-trial voxel-representational stability in the right angular gyrus, right orbitofrontal cortex, and right inferior frontal gyrus. RDM model fits and mnemonic stability were predicted by total CA23 volumes. Trial-by-trial retrieval stability within the right orbitofrontal cortex and right inferior frontal gyrus predicted autobiographical memory performance, providing the first direct neural correlate between hippocampal dysfunction, altered mnemonic representations, and amnesia. The results demonstrate that hippocampal-mediated amnesia results from impaired mnemonic representational distinctiveness and stability within the AMN.

neuroscience↗

bioRxiv2024

Evaluation of Brain Age as a Specific Marker of Brain Health

Brain age is widely regarded as a powerful marker of general brain health. Brain age models are typically trained on large datasets to predict chronological age, which may offer advantages in predicting specific health outcomes, much like the success of finetuning large language models for specific applications. However, it is also well-accepted that machine learning models trained to directly predict specific outcomes (i.e., direct models) often outperform those trained on surrogate objectives. Therefore, despite their much larger training data, it is unclear whether brain age models outperform direct models in predicting specific brain health outcomes. Here, we compare large-scale brain age models (pretrained on 53,542 participants) and direct models for predicting specific health outcomes related to Alzheimers Disease (AD) dementia. Using anatomical T1 scans from three continents (N = 1,848), we find that summarizing brain age with a single scalar (i.e., brain age gap) led to poor prediction performance. Using higher-dimensional intermediate representations of brain age models led to better prediction, but was still worse than direct models without finetuning. Using intermediate representations of finetuned brain age models was necessary to achieve similar performance as direct models. Overall, our results do not discount brain age as a useful marker of general brain health, but suggest that using chronological age as a pretraining target might be suboptimal for predicting specific health outcomes.

neuroscience↗

bioRxiv2024

WMH-DualTasker: A Dual-Task Deep Learning Model with Self-supervised Consistency for Automated Segmentation and Visual Rating of White matter Hyperintensities - a Multicentre study

BackgroundWhite matter hyperintensities (WMH) are neuroimaging markers linked to an elevated risk of cognitive decline. WMH severity is typically assessed via visual rating scales and through volumetric segmentation. While visual rating scales are commonly used in clinical practice, they offer limited descriptive power. In contrast, supervised volumetric segmentation requires manually annotated masks, which is labor-intensive and challenging to scale for large studies. Therefore, our goal was to develop an automated deep learning model that can provide accurate and holistic quantification of WMH severity with minimal supervision. MethodsWe developed WMH-DualTasker, a deep learning model that simultaneously performs voxel-wise segmentation and visual rating score prediction. The model employs self-supervised learning with transformation-invariant consistency constraints, using WMH visual ratings (ARWMC scale, range 0-30) from clinical settings as the sole supervisory signal. Additionally, we assessed its clinical utility by applying it to identify individuals with mild cognitive impairment (MCI) and to predict dementia conversion. FindingsThe volumetric quantification performance of WMH-DualTasker was either superior to or on par with existing supervised methods, as demonstrated on the MICCAI-WMH dataset (N=60, Dice=0.602) and the SINGER dataset (N=64, Dice=0.608). Furthermore, the model exhibited strong agreement with clinical visual rating scales on an external dataset (SINGER, MAE=1.880, K=0.77). Importantly, WMH severity metrics derived from WMH-DualTasker improved predictive performance beyond conventional clinical features for MCI classification (AUC=0.718, p<0.001), MCI conversion prediction (AUC=0.652, p<0.001) using the ADNI dataset. InterpretationsWMH-DualTasker substantially reduces the reliance on labor-intensive manual annotations, facilitating more efficient and scalable quantification of WMH severity in large-scale population studies. This innovative approach has the potential to advance preventive and precision medicine by enhancing the assessment and management of vascular cognitive impairment associated with WMH. Code and model weights are publicly available at https://github.com/hzlab/WMH-DualTasker.

neuroscience↗

bioRxiv2024

Additive effects of cerebrovascular disease functional connectome phenotype and plasma p-tau181 on longitudinal neurodegeneration and cognitive outcomes

INTRODUCTIONWe investigated the effects of multiple cerebrovascular disease (CeVD) neuroimaging markers on brain functional connectivity (FC), and how such CeVD-related FC changes interact with plasma p-tau181 (Alzheimers disease (AD) marker) to influence downstream neurodegeneration and cognitive changes. METHODSMultivariate associations between four CeVD markers and whole-brain FC in 529 participants across the dementia spectrum were examined using partial least squares correlation. Interactive effects of CeVD-related FC patterns and p-tau181 on longitudinal grey matter volume and cognitive changes were investigated using linear mixed-effects models. RESULTSWe identified a brain FC phenotype associated with high CeVD burden across all markers. Further, expression of this general CeVD-related FC phenotype and p-tau181 contributed additively, but not synergistically, to baseline and longitudinal grey matter volumes and cognitive changes. DISCUSSIONOur findings suggest that CeVD exerts global effects on the brain connectome and highlight the additive nature of AD and CeVD on neurodegeneration and cognition.

neuroscience↗

bioRxiv2024

Test Retest Reliability of Meta Analytic Networks During Naturalistic Viewing

Functional connectivity analyses have given considerable insights into human brain function and organization. As research moves towards clinical application, test-retest reliability has become a main focus of the field. So far, the majority of studies have relied on resting-state paradigms to examine brain connectivity, based on its low demand and ease of implementation. However, the reliability of resting-state measures is mostly moderate, potentially due to its unconstrained nature. Recently, naturalistic viewing paradigms have gained popularity because they probe the human brain under more ecologically valid conditions, thereby possibly increasing reliability. Therefore, we here compared the reliability of graph metrics extracted from resting-state and naturalistic viewing in functional networks, across two sessions. We show that naturalistic viewing can increase reliability over resting-state, but that its effect varies between stimuli and networks. Furthermore, we demonstrate that the effect of naturalistic viewing differs between two cohorts with Asian and European cultural backgrounds. Taken together, our study encourages the use of naturalistic viewing to increase reliability, but emphasizes the need to carefully select the appropriate stimulus and network for the respective research question.

neuroscience↗

bioRxiv2024

MRI economics: Balancing sample size and scan duration in brain wide association studies

A pervasive dilemma in brain-wide association studies (BWAS) is whether to prioritize functional MRI (fMRI) scan time or sample size. We derive a theoretical model showing that individual-level phenotypic prediction accuracy increases with sample size and total scan duration (sample size x scan time per participant). The model explains empirical prediction accuracies extremely well across 76 phenotypes from nine resting-fMRI and task-fMRI datasets (R2 = 0.89), spanning a wide range of scanners, acquisitions, racial groups, disorders and ages. For scans [≤]20 mins, prediction accuracy increases linearly with the logarithm of total scan duration, suggesting interchangeability of sample size and scan time. However, sample size is ultimately more important than scan time in determining prediction accuracy. Nevertheless, when accounting for overhead costs associated with each participant (e.g., recruitment costs), to boost prediction accuracy, longer scans can yield substantial cost savings over larger sample size. To achieve high prediction performance, 10-min scans are highly cost inefficient. In most scenarios, the optimal scan time is [≥]20 mins. On average, 30-min scans are the most cost-effective, yielding 22% cost savings over 10-min scans. Overshooting is cheaper than undershooting the optimal scan time, so we recommend aiming for [≥]30 mins. Compared with resting-state whole-brain BWAS, the most cost-effective scan time is shorter for task-fMRI and longer for subcortical-cortical BWAS. Standard power calculations maximize sample size at the expense of scan time. Our study demonstrates that optimizing both sample size and scan time can boost prediction power while cutting costs. Our empirically informed reference is available for future study planning: WEB_APPLICATION_LINK

neuroscience↗

bioRxiv2024

DeepResBat: deep residual batch harmonization accounting for covariate distribution differences

Pooling MRI data from multiple datasets requires harmonization to reduce undesired inter-site variabilities, while preserving effects of biological variables (or covariates). The popular harmonization approach ComBat uses a mixed effect regression framework that explicitly accounts for covariate distribution differences across datasets. There is also significant interest in developing harmonization approaches based on deep neural networks (DNNs), such as conditional variational autoencoder (cVAE). However, current DNN approaches do not explicitly account for covariate distribution differences across datasets. Here, we provide mathematical results, suggesting that not accounting for covariates can lead to suboptimal harmonization. We propose two DNN-based covariate-aware harmonization approaches: covariate VAE (coVAE) and DeepResBat. The coVAE approach is a natural extension of cVAE by concatenating covariates and site information with site- and covariate-invariant latent representations. DeepResBat adopts a residual framework inspired by ComBat. DeepResBat first removes the effects of covariates with nonlinear regression trees, followed by eliminating site differences with cVAE. Finally, covariate effects are added back to the harmonized residuals. Using three datasets from three continents with a total of 2787 participants and 10085 anatomical T1 scans, we find that DeepResBat and coVAE outperformed ComBat, CovBat and cVAE in terms of removing dataset differences, while enhancing biological effects of interest. However, coVAE hallucinates spurious associations between anatomical MRI and covariates even when no association exists. Future studies proposing DNN-based harmonization approaches should be aware of this false positive pitfall. Overall, our results suggest that DeepResBat is an effective deep learning alternative to ComBat. Code for DeepResBat can be found here: https://github.com/ThomasYeoLab/CBIG/tree/master/stable_projects/harmonization/An2024_DeepResBat.

neuroscience↗

bioRxiv2023

Investigating the validity and interpretability of longitudinal brain age underlying cognition in Asian children and older adults

Brain age has emerged as a powerful tool to understand neuroanatomical aging and its link to health outcomes like cognition. However, there remains a lack of studies investigating the rate of brain aging and its relationship to cognition. Furthermore, most brain age models are trained and tested on cross-sectional data from primarily Caucasian, adult participants. It is thus unclear how well these models generalize to non-Caucasian participants, especially children. Here, we tested a previously published deep learning model on Singaporean elderly participants (55 - 88 years old) and children (4 - 11 years old). We found that the model directly generalized to the elderly participants, but model finetuning was necessary for children. After finetuning, we found that the rate of change in brain age gap was associated with future executive function performance in both elderly participants and children. We further found that lateral ventricles and frontal areas contributed to brain age prediction in elderly participants, while white matter and posterior brain regions were more important in predicting brain age of children. Taken together, our results suggest that there is potential for generalizing brain age models to diverse populations. Moreover, the longitudinal change in brain age gap reflects developing and aging processes in the brain, relating to future cognitive function.

neuroscience↗

bioRxiv2023

In-vivo whole-cortex estimation of excitation-inhibition ratio indexes cortical maturation and cognitive ability in youth

A balanced excitation-inhibition ratio (E/I ratio) is critical for healthy brain function. Normative development of cortex-wide E/I ratio remains unknown. Here we non-invasively estimate a putative marker of whole-cortex E/I ratio by fitting a large-scale biophysically-plausible circuit model to resting-state functional MRI (fMRI) data. We first confirm that our model generates realistic brain dynamics in the Human Connectome Project. Next, we show that the estimated E/I ratio marker is sensitive to the GABA-agonist benzodiazepine alprazolam during fMRI. Alprazolam-induced E/I changes are spatially consistent with positron emission tomography measurement of benzodiazepine receptor density. We then investigate the relationship between the E/I ratio marker and neurodevelopment. We find that the E/I ratio marker declines heterogeneously across the cerebral cortex during youth, with the greatest reduction occurring in sensorimotor systems relative to association systems. Importantly, among children with the same chronological age, a lower E/I ratio marker (especially in association cortex) is linked to better cognitive performance. This result is replicated across North American (8.2 to 23.0 years old) and Asian (7.2 to 7.9 years old) cohorts, suggesting that a more mature E/I ratio indexes improved cognition during normative development. Overall, our findings open the door to studying how disrupted E/I trajectories may lead to cognitive dysfunction in psychopathology that emerges during youth. SignificanceHealthy brain function requires a delicate balance of neural excitation (E) and inhibition (I). In animals, this balance - the E/I ratio - is known to decrease with the maturation of inhibitory circuitry during healthy development. However, in humans, the normative development of cortex-wide E/I ratio remains unclear. Here, we use a biophysical model and non-invasive brain scans to estimate a marker of E/I ratio. Spatial changes in our E/I ratio marker are consistent with a drug that decreases E/I ratio. We also find that our cortex-wide E/I ratio marker decreases during development. Furthermore, North American and Asian children with lower E/I ratio, especially in higher-order cortex, have better cognitive performance. Overall, E/I ratio is a potential index of healthy neurocognitive development.

neuroscience↗

bioRxiv2022

Homotopic local-global parcellation of the human cerebral cortex from resting-state functional connectivity

Resting-state fMRI is commonly used to derive brain parcellations, which are widely used for dimensionality reduction and interpreting human neuroscience studies. We previously developed a model that integrates local and global approaches for estimating areal-level cortical parcellations. The resulting local-global parcellations are often referred to as the Schaefer parcellations. However, the lack of homotopic correspondence between left and right Schaefer parcels has limited their use for brain lateralization studies. Here, we extend our previous model to derive homotopic areal-level parcellations. Using resting-fMRI and task-fMRI across diverse scanners, acquisition protocols, preprocessing and demographics, we show that the resulting homotopic parcellations are as homogeneous as the Schaefer parcellations, while being more homogeneous than five publicly available parcellations. Furthermore, weaker correlations between homotopic parcels are associated with greater lateralization in resting network organization, as well as lateralization in language and motor task activation. Finally, the homotopic parcellations agree with the boundaries of a number of cortical areas estimated from histology and visuotopic fMRI, while capturing sub-areal (e.g., somatotopic and visuotopic) features. Overall, these results suggest that the homotopic local- global parcellations represent neurobiologically meaningful subdivisions of the human cerebral cortex and will be a useful resource for future studies. Multi-resolution parcellations estimated from 1479 participants are publicly available (GITHUB_LINK).

neuroscience↗

bioRxiv2022

Restoration of functional PAX6 in aniridia patient iPSC-derived ocular tissue models using repurposed nonsense suppression drugs

Aniridia is a rare, pan-ocular disease causing severe sight loss, with only symptomatic intervention offered to patients. Approximately 40% of aniridia patients present with heterozygous nonsense variants in PAX6, resulting in haploinsufficiency. Translational readthrough inducing compounds (TRIDs) have the ability to weaken the recognition of in-frame premature stop codons (PTCs), permitting full-length protein to be translated. We have established induced pluripotent stem cell (iPSC)-derived 3D optic cups and 2D limbal epithelial stem cell (LESC) models from an aniridia patient with a prevalent PAX6 nonsense mutation. Both in vitro models show reduced PAX6 protein levels, mimicking the disease. Repurposed TRIDs amlexanox and 2,6-diaminopurine (DAP), and positive control compounds ataluren and G418 were tested for their efficiency. Amlexanox was identified as the most promising TRID, increasing full-length PAX6 levels in both models, and rescuing the disease phenotype through normalization of VSX2 and cell proliferation in the optic cups and reduction of ABCG2 protein and SOX10 expression in LESC. This study highlights the significance of patient iPSC-derived cells as a new model system for aniridia and proposes amlexanox as a new putative treatment for nonsense-mediated aniridia.

genetics↗

bioRxiv2022

Comparison of individualized behavioral predictions across anatomical, diffusion and functional connectivity MRI

A fundamental goal across the neurosciences is the characterization of relationships linking brain anatomy, functioning, and behavior. Although various MRI modalities have been developed to probe these relationships, direct comparisons of their ability to predict behavior have been lacking. Here, we compared the ability of anatomical T1, diffusion and functional MRI (fMRI) to predict behavior at an individual level. Cortical thickness, area and volume were extracted from anatomical T1 images. Diffusion Tensor Imaging (DTI) and approximate Neurite Orientation Dispersion and Density Imaging (NODDI) models were fitted to the diffusion images. The resulting metrics were projected to the Tract-Based Spatial Statistics (TBSS) skeleton. We also ran probabilistic tractography for the diffusion images, from which we extracted the stream count, average stream length, and the average of each DTI and NODDI metric across tracts connecting each pair of brain regions. Functional connectivity (FC) was extracted from both task and resting-state fMRI. Individualized prediction of a wide range of behavioral measures were performed using kernel ridge regression, linear ridge regression and elastic net regression. Consistency of the results were investigated with the Human Connectome Project (HCP) and Adolescent Brain Cognitive Development (ABCD) datasets. In both datasets, FC-based models gave the best prediction performance, regardless of regression model or behavioral measure. This was especially true for the cognitive domain. Furthermore, all modalities were able to predict cognition better than other behavioral domains. Combining all modalities improved prediction of cognition, but not other behavioral domains. Finally, across all behaviors, combining resting and task FC yielded prediction performance similar to combining all modalities. Overall, our study suggests that in the case of healthy children and young adults, behaviorally-relevant information in T1 and diffusion features might reflect a subset of the variance captured by FC. HighlightsO_LIFC predicts behavior better than anatomical and diffusion features C_LIO_LICognition is predicted better than other behavioral domains regardless of modality C_LIO_LICombining resting & task FC improves prediction as much as combining all modalities C_LIO_LIFindings were replicated over 3 regression models and 2 datasets C_LI

neuroscience↗

bioRxiv2022

Goal-specific brain MRI harmonization

There is significant interest in pooling magnetic resonance image (MRI) data from multiple datasets to enable mega-analysis. Harmonization is typically performed to reduce heterogeneity when pooling MRI data across datasets. Most MRI harmonization algorithms do not explicitly consider downstream application performance during harmonization. However, the choice of downstream application might influence what might be considered as study-specific confounds. Therefore, ignoring downstream applications during harmonization might potentially limit downstream performance. Here we propose a goal-specific harmonization framework that utilizes downstream application performance to regularize the harmonization procedure. Our framework can be integrated with a wide variety of harmonization models based on deep neural networks, such as the recently proposed conditional variational autoencoder (cVAE) harmonization model. Three datasets from three different continents with a total of 2787 participants and 10085 anatomical T1 scans were used for evaluation. We found that cVAE removed more dataset differences than the widely used ComBat model, but at the expense of removing desirable biological information as measured by downstream prediction of mini mental state examination (MMSE) scores and clinical diagnoses. On the other hand, our goal-specific cVAE (gcVAE) was able to remove as much dataset differences as cVAE, while improving downstream cross-sectional prediction of MMSE scores and clinical diagnoses.

neuroscience↗