In major depressive disorder (MDD), episodic memory is compromised and linked to hippocampal dysfunction, yet the underlying neural circuitry remains poorly understood. We examined the circuits supporting episodic memory and assessed its MDD-associated anatomical connectivity fingerprints. Anatomical connectivity was assessed using magnetic resonance imaging (MRI) relaxometry-derived indices of regional volume and myelin content from 1339 community-dwelling adults (mean age 61 ± 9 y; 671 females). To obtain an unbiased definition of episodic memory circuits, we estimated whole-brain anatomical covariance patterns originating from hippocampal regions that, in an independent sample, showed differential MDD-related neural activity patterns during a functional MRI episodic memory task. The morphometry-based episodic memory circuit encompassing frontal, parietal and temporal cortex, amygdala, pallidum, and thalamus (all |r| > 0.106, P < .05), showed similar covariance patterns for the anterior and posterior hippocampal regions. Conversely, myelin content-defined circuits showed inverse covariance patterns between cortical and subcortical structures, yet convergent for the two hippocampal regions. Connectivity with the orbital gyri, pallidum, insula, and, in particular, the thalamic nuclei was discriminative (all |Z| > 1.96, all P < .05, two-tailed) for individuals with a lifetime MDD diagnosis (n = 700, mean age 59 ± 9 y; 432 females) compared to those without (n = 639, mean age 63 ± 10 y; 239 females). Our findings provide evidence of altered brain circuits underlying depression-related dysfunctional episodic memory. Altered connectivity in these circuits suggests roles for multiple pathways in the maintenance and aetiology of depression and represents a potential target for therapeutic interventions.

Abstract Multi-parameter mapping (MPM) magnetic resonance imaging (MRI) provides parameters sensitive to myelin, iron and water. Conventional analyses treat these parameters individually or via pairwise correlations. We introduce O-information ( $$\Omega$$ Ω ) as a proof-of-principle higher-order interaction framework to quantify how the three myelin-sensitive parameters (magnetisation transfer, longitudinal relaxation rate $${R}_{1}$$ R 1 and proton density) are interrelated beyond pairwise dependencies. We compute $$\Omega$$ Ω from their joint distribution across cortical grey matter, subcortical grey matter and white-matter bundles and ask how $$\Omega$$ Ω is modulated by the iron-sensitive transverse relaxation rate $${R}_{2}^{*}$$ R 2 ∗ in grey matter and by fibre architecture (neurite density and orientation dispersion) in white matter. In 22 healthy adults, $$\Omega$$ Ω separates tissue classes (cortex near-balanced, subcortex mildly synergistic, white matter strongly synergistic), consistent in 21 of 22 participants. At matched voxel count per region, grey-matter $$\Omega$$ Ω declines with $${R}_{2}^{*}$$ R 2 ∗ and changes sign from redundancy to synergy within the physiological $${R}_{2}^{*}$$ R 2 ∗ range, an effect carried by cortex. A minimal two-compartment forward model [1] ( $${R}_{2}^{*}\propto \left[{\text{Fe}}\right]$$ R 2 ∗ ∝ Fe , following Langkammer et al. 257:455–462, 2010 ) generates the same sign change using constants taken from the literature, but places it at systematically lower $${R}_{2}^{*}$$ R 2 ∗ . Fitting a single effective iron- $${R}_{1}$$ R 1 coupling closes the gap, at a value below the literature estimate. In white matter, $$\Omega$$ Ω is only weakly coupled to $${R}_{2}^{*}$$ R 2 ∗ ( $$r=+0.007$$ r = + 0.007 , $$p=0.82$$ p = 0.82 ), well below the grey-matter coupling. Neurite density, not orientation dispersion, is the dominant inter-bundle predictor. Because NODDI and $$\Omega$$ Ω are both derived in part from shared tissue properties, the association does not establish an independent causal link. O-information thus quantifies how myelin-sensitive parameters combine or diverge across tissue compartments, providing a higher-order axis not captured by pairwise analyses.

Despite advances in neuromelanin-sensitive brain imaging and a plethora of software solutions, the reliable non-invasive delineation of the locus coeruleus (LC) in the human brainstem remains challenging. We sought to evaluate the spatial accuracy and consistency of atlas- and probabilistic spatial prior-based LC delineation. We acquired neuromelanin-sensitive 3 T MRI data in healthy volunteers (n = 24; mean age 40.0 ± 16.8 years; 42% female). Manual labelling by 9 raters performed twice provided the basis for individual- and group-level comparisons, showing moderate inter-rater agreement (mean Dice = 0.7). For the atlas-based labelling, we tested seven open-access LC atlases, a consensus reference representing the atlases' overlap and the averaged manual labelling. Open-access atlases demonstrated low spatial concordance (Dice = 0.2-0.4), while the averaged manual labelling atlas had higher spatial overlap (Dice = 0.6). Probabilistic delineation using spatial priors showed the strongest voxel-wise similarity with manual labelling (r = 0.3) when the averaged manual labelling atlas was used as prior. Principal component analysis confirmed the greater spatial compactness for atlas-based labelling. Atlas-based labelling using the averaged manual labelling atlas in an extended 3 T cohort (n = 2393, mean age 58.44 ± 13.75 years) identified that tissue myelin and iron declined continuously from early adulthood, while free tissue water increased - a neurobiological trend robust to atlas choice. LC volume showed an inverted-U trajectory. Our results highlight the potential of atlas-based labelling for LC identification and demonstrate its sensitivity to physiologically grounded aging processes, suggesting that harmonised validation strategies and context-sensitive approaches can improve reliability.

White matter hyperintensities (WMHs) are the imaging hallmark of cerebral small vessel disease (SVD), yet their microstructural composition, spatial heterogeneity, and relationship to diffuse normal-appearing white matter (NAWM) damage and cardiovascular risk remain incompletely defined. In 363 community-dwelling adults from the BrainLaus cohort (mean age 55.5 years; range 19.8-89.4; 48.8% male), we combined quantitative relaxometry (MTsat, R1, R2*) and diffusion-derived metrics (FA, MD, NODDI, g-ratio). WMHs were automatically segmented on FLAIR and microstructure was quantified across lobes, white matter compartments, and geodesic layers extending from the WMH core into surrounding NAWM. Multivariate organisation was assessed using principal component analysis, and associations with cardiovascular risk factors were tested using partial least squares. Our analysis revealed demyelination, axonal loss, and extracellular fluid accumulation, particularly in periventricular regions. Layer-specific profiles showed a centrifugal gradient, with myelin loss and oedema at the core and axonal alteration in surrounding tissue. These signatures were associated with age-related cardiovascular risk factors, including higher blood pressure, bioimpedance, and lower haemoglobin levels. WMHs index the endpoint of a broader, spatially structured white matter injury process that extends into NAWM, is regionally concentrated in periventricular tissue, and covaries with systemic vascular/metabolic factors. These findings support sustained vascular risk management to mitigate CSVD-related white matter degeneration.

White matter hyperintensities (WMH) are a hallmark of cerebral small vessel disease, a highly prevalent condition in aging, making their accurate detection in large-scale magnetic resonance imaging (MRI) studies critically important. Currently available semi-automated and automated segmentation methods are frequently limited by sensitivity to variations in imaging conditions and computational constraints, restricting their applicability across diverse acquisition settings. We present WHITE-Net, a deep learning-based framework for automated WMH segmentation built on a 3D ResUNet architecture and trained on multisite MRI data encompassing substantial variability in scanner vendors, acquisition protocols, spatial resolution, and image characteristics. Evaluated across three independent datasets - BrainLaus, ADNI, and WMH Challenge, WHITE-Net achieves high segmentation accuracy and ranks among the top-performing methods across datasets, enabling a comprehensive assessment of generalisability in varied imaging environments. WHITE-Net maintains stable performance in a wide spectrum of lesion loads and demonstrates a favorable balance between precision and recall, effectively limiting false positive detections - a common limitation of existing approaches. Beyond accuracy, WHITE-Net requires no parameter tuning and offers fast inference times, making it well-suited for deployment in large-scale neuroimaging studies. The obtained results highlight the value of training on diverse multisite data for improving generalisability and position WHITE-Net as a reliable, scalable tool for automated WMH segmentation in computational anatomy research. ### Competing Interest Statement The authors have declared no competing interest. ### Clinical Protocols ### Funding Statement Funding is supported by the Swiss National Science Foundation (project grants Nr. 213595,32003B\_135679, 32003B\_159780, 324730\_192755 and CRSK-3\_190185), ERA_NET iSEE and BrainTree projects. LREN is very grateful to the Roger De Spoelberch and Partridge Foundations for their generous financial support. ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: Ethics Commission of Canton de Vaud gave ethical approval for this work I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes Data of CoLaus|PsyCoLaus study used in this article cannot be fully shared as they contain potentially sensitive personal information on participants. According to the Ethics Committee for Research of the Canton of Vaud, sharing these data would be a violation of the Swiss legislation with respect to privacy protection. However, coded individual-level data that do not allow researchers to identify participants are available upon request to researchers who meet the criteria for data sharing of the CoLaus|PsyCoLaus Datacenter (CHUV, Lausanne, Switzerland). Any researcher affiliated to a public or private research institution who complies with the CoLaus|PsyCoLaus standards can submit a research application to research.colaus{at}chuv.ch or research.psycolaus{at}chuv.ch. Proposals will be evaluated by the Scientific Committee (SC) of the CoLaus|PsyCoLaus studies. Detailed instructions for gaining access to the CoLaus|PsyCoLaus data used in this study are available at www.colaus-psycolaus.ch/professionals/how-to-collaborate/. Data from WMH challenge are freely available at https://doi.org/10.34894/aecrsd (H. Kuijf et al., 2022). Code for the segmentation tool described in this article, WHITE-Net, is available on GitHub (https://github.com/cathalacamille/WHITE-Net). Swiss National Science Foundation, https://ror.org/00yjd3n13, 213595, 32003B_135679, 32003B_159780, 324730_192755, CRSK-3_190185 ERA-NET NEURON JTC-2020 iSEE ERA-NET NEURON JTC-2023 ELSA Agence Nationale de la Recherche, ANR-23-IACL-0008, ANR-10-IAIHU-06 EU Horizon Europe Framework, 101136607 Inserm, AAP-2023-MSDR-341011

There is much controversy about the spatial heterogeneity and microstructural characterisation of white matter hyperintensities (WMHs), the neuroimaging hallmark of cerebral small vessel disease. To address this knowledge gap, we analysed relaxometry and diffusion-weighted magnetic resonance imaging (MRI) together with cardiovascular risk data in a community-dwelling adult cohort (202 participants, mean age: 69.2 years, range: 53.3-85.4 years, 110 females). Following automated WMH detection, we compared MRI-derived metrics of tissue myelination, iron content, axonal density, and extracellular water, between WMH and normal-appearing white matter (NAWM). Principal component analysis demonstrated a pattern of demyelination, axonal loss, and extracellular fluid accumulation, particularly pronounced in periventricular regions. The layer-specific profiles of theWMHs revealed a centrifugal gradient of white matter vulnerability, with myelin loss and extracellular water accumulation, interpreted as oedema, at the core, and ongoing axonal reorganization and demyelination in the surrounding tissue. The obtained signatures of white matter pathology were intimately linked to the aging-related increase in cardiovascular risk. High blood pressure and impaired glucose regulation, together with reduced cholesterol and hemoglobin levels, emerged as key contributors. Our findings suggest that WMHs represent the radiologically visible endpoint of a more widespread white matter damage in cerebral small vessel disease. This underscores the pressing need for early detection and consequent treatment of cardiovascular risk factors. ### Competing Interest Statement The authors have declared no competing interest. Swiss National Science Foundation, 213595, 32003B_135679, 32003B_159780, 324730_192755, CRSK-3_190185, 320030_184784 Innosuisse – Swiss Innovation Agency, https://ror.org/05a2bhn71, Flagship Swiss brAInHealth project ERA_NET NEURON, JTC2020 Horizon Europe grant, 1010953844:01 Horizon 2020, 871643-MORPHEMIC Fondation Roger de Spoelberch

Cerebral small vessel disease, a major cause of aging-related stroke and cognitive decline, often manifests as white matter hyperintensities (WMH) on MRI1–3. WMHs vary in location and tissue microstructure, reflecting underlying pathology such as demyelination and axonal loss. Cardiovascular risk factors (CVR), including elevated systolic blood pressure (SBP) and heart rate, are linked to WMH development4. This study explores the relationship between longitudinal CVR and WMH characteristics, including quantitative MRI (qMRI) analyses of microstructural alterations. Data were obtained from 539 participants in the BrainLaus study (CoLaus|PsycoLaus cohort, Lausanne, Switzerland). 153 participants (86 females) were selected on the basis of at least three CVR measurements, with a mean follow-up duration of 14.45 years. CVR factors included SBP, diastolic blood pressure (DBP), body mass index (BMI), bioimpedance, waist-to-hip ratio (WHR), heart rate, cholesterol (HDL, LDL, total), triglycerides, glucose, and insulin. Imaging was performed on a 3T Magnetom Prisma using FLAIR, T1-weighted MPRAGE, multi-echo FLASH, and DWI. WMH were segmented using a ResUNet-based algorithm 5, and quantitative maps (MTsat, R2*, R1) were generated 6. CVR over time was modeled using splines, and area under the curve (AUC) was calculated for each participant, normalized by total duration. Statistical analysis included linear models with sex interaction terms, corrected for false discovery rate (FDR). Figure 1 shows CVR AUC over time, highlighting inter-individual variation. Figure 2 demonstrates that increased CVR (SBP, LDL, glucose, WHR, bioimpedance, total cholesterol) was associated with larger WMH volumes, with stronger effects in women for SBP, LDL, total cholesterol, and triglycerides. Figure 3 shows that CVR impacts WMH microstructure, with cholesterol-related risk factors linked to higher water content (MD, ISOVF) in women, and hypertension/heart rate linked to lower myelin content (MT, R1) in men. Men showed increased g-ratio indicating myelin thickness reduction with higher CVR. The impact of CVR on WMH differs by sex, with women showing greater effects on water content and men more on myelin integrity. These findings suggest that sex-specific approaches may be needed for understanding and targeting cardiovascular risk factors to preserve white matter health and prevent cognitive decline.

The study aimed to assess the discriminative capacity of a machine learning algorithm in distinguishing between individuals with Major Depressive Disorder and healthy controls based on a dataset collected during the performance of a Stroop Color and Word Test combined with an n-back component in functional magnetic resonance imaging. A total of 50 participants were recruited, including 24 patients with depression and 26 healthy controls. The analysis employed a multivariate linear model, which identified two principal components characterized by their eigenvalues. The key finding of our study highlights the distinct contribution of eigenvalues, as represented in the principal components, to brain signatures with a strong capacity to differentiate between the two diagnostic groups examined for depression and healthy controls. Moreover, the results present a fresh network-level perspective, emphasizing the intricate interactions among different brain networks in major depression disorder. These findings support prior research indicating disruptions in sensory processing, cognitive control, and emotional regulation in Major Depressive Disorder. The results provide a novel, network-level perspective on these alterations, emphasizing the intricate interplay between sensory, cognitive, and emotional processes. Understanding these network dynamics may offer valuable insights into the neural mechanisms of Major Depressive Disorder and inform targeted interventions aimed at restoring functional connectivity and improving symptom management.

This paper explores how multivariate methods can be applied to machine learning in mental health research. These techniques help uncover relationships between variables like symptoms, behavior, brain function, and genetics, offering insights into personalized interventions. We introduce key multivariate approaches—such as PCA, factor analysis, and partial least squares—alongside their theoretical foundations, benefits, and limitations. Real- world examples illustrate their use in predicting treatment outcomes and identifying risk factors. Finally, we discuss methodological challenges, emphasizing the importance of large datasets and the integration of model-based and AI-driven approaches in computational neuroscience.

Despite major progress in understanding the impact of the triplicated chromosome 21 on the brain and behaviour in Down syndrome, our knowledge of the underlying neurobiology in humans is still limited. We sought to address some of the pertinent questions about the drivers of brain structure differences and their associations with cognitive function in Down syndrome. To this aim, in a pilot magnetic resonance imaging (MRI) study, we monitored brain anatomy in individuals with Down syndrome receiving pulsatile gonadotropin-releasing hormone (GnRH) therapy over 6 months in comparison with typically developed age- and sex-matched healthy controls. We analysed cross-sectional (Down syndrome/healthy controls n  = 11/27; Down syndrome-2 females/9 males, age 26.7 ± 5.0 years old; healthy controls-8 females/19 males, age 24.1 ± 2.5 years old) and longitudinal (Down syndrome/healthy controls n  = 8/13; Down syndrome-1 female/7 males, age 26.4 ± 5.3 years old; healthy controls-4 females/9 males, 24.7 ± 2.2 years old) relaxometry and diffusion-weighted MRI data alongside standard cognitive assessment. The statistical tests looked for cross-sectional baseline differences and for differential changes over time between Down syndrome and healthy controls. The post hoc analysis confined to the Down syndrome group, tested for potential time-dependent interactions between individuals' overall cognitive performance and associated brain anatomy changes. The brain MRI statistical analyses covered both grey and white matter regions across the whole brain allowing for investigation of regional volume, macromolecular/myelin and iron content, additionally to diffusion tensor and neurite orientation and dispersion density characterization across major white matter tracts. The cross-sectional analysis showed reduced frontal, temporal and cerebellar volumes in Down syndrome with only the cerebellar differences remaining significant after adjustment for the presence of microcephaly (P family-wise-corrected < 0.05). The volume reductions were paralleled by decreased cortical and subcortical macromolecular/myelin content confined to the cortical motor system, thalamus and basal ganglia (P family-wise-corrected < 0.05). All major white matter tracts showed a ubiquitous mean diffusivity and intracellular volume fraction reduction contrasted with no differences in magnetization transfer saturation metrics (P family-wise-corrected < 0.05). Compared with healthy controls over the same period, Down syndrome individuals under GnRH therapy showed cognitive improvement (Montreal Cognitive Assessment from 11.4 ± 5.5 to 15.1 ± 5.6; P < 0.01) on the background of stability of the observed differential neuroanatomical patterns. Despite the lack of adequate Down syndrome control group, we interpret the obtained cross-sectional and longitudinal findings in young adults as evidence for predominant neurodevelopmental neuronal loss due to dysfunctional neurogenesis without signs for short-term myelin loss.

Given the association of Epstein-Barr virus (EBV) with subjective perception of fatigue and demyelination in clinical conditions, the question about potential subclinical effects in the adult general population remains open. We investigate the association between individuals' EBV immune response and perceived fatigue in a community dwelling cohort (n = 864, age 62 +/- 10 years old; 49% women) while monitoring brain tissue properties. Fatigue levels are assessed with the established fatigue severity scale, the EBNA-1 and VCA p18 immunoglobulin G (IgG) chronic response - with multiplex serology and the estimates of local brain volume, myelin content, and axonal density- using relaxometry- and multi-shell diffusion-based magnetic resonance imaging (MRI). In our analysis we adjust for the effects of demographic and cardiovascular risk factors, sleep apnea, depression, and polygenic risk score for multiple sclerosis. We demonstrate that EBNA-1 IgG levels are positively associated with perceived levels of fatigue, whilst VCA p18 IgG levels show a positive correlation with myelin content and a negative one with an estimate of axonal g-ratio in male participants. In the context of EBVs immune response, the polygenic risk for multiple sclerosis is not associated with increased fatigue levels, brain myelination or atrophy. Our findings bring empirical evidence about the potential role of EBVs chronic immune response in perceived fatigue and hint towards a protective role of myelination specific for men. They underscore the added value of advanced assessment of brain tissue microstructure in uncovering the mechanisms behind frequent fatigue complaints associated with EBV infection and multiple sclerosis.

In face of cumulating evidence about the impact of human-induced environmental changes on mental health and behavior, our understanding of the main effects and interactions between environmental factors - i.e., the exposome and the brain - is still limited. We seek to fill this knowledge gap by leveraging georeferenced large-scale brain imaging and psychometry data from the adult community-dwelling population (n = 2672; mean age 63 ± 10 years). For monitoring brain anatomy, we extract morphometry features from a nested subset of the cohort (n = 944) with magnetic resonance imaging. Using an iterative analytical strategy testing the moderator role of geospatially encoded exposome factors on the association between brain anatomy and psychometry, we demonstrate that individuals' anxiety state and psychosocial functioning are among the mental health characteristics showing associations with the urban exposome. The clusters of higher anxiety state and lower current psychosocial functioning coincide spatially with a lower vegetation density and higher air pollution. The univariate multiscale geographically weighted regression identifies the spatial scale of associations between individuals' levels of anxiety state, psychosocial functioning, and overall cognition with vegetation density, air pollution and structures of the limbic network. Moreover, the multiscale geographically weighted regression interaction model reveals spatially confined exposome features with moderating effect on the brain-psychometry/cognitive performance relationships. Our original findings testing the role of exposome factors on brain and behavior at the individual level, underscore the role of environmental and spatial context in moderating brain-behavior dynamics across the adult lifespan.

Despite major advances, our understanding of the neurobiology of life course socioeconomic conditions is still scarce. This study aimed to provide insight into the pathways linking socioeconomic exposures – household income, last-known occupational position, and life course socioeconomic trajectories – with brain microstructure and cognitive performance in middle to late adulthood. We assessed socioeconomic conditions alongside quantitative relaxometry and diffusion-weighted magnetic resonance imaging indicators of brain tissue microstructure, and cognitive performance in a sample of community-dwelling men and women (N=751, aged 50-91 years). We adjusted the applied regression analyses and structural equation models for the linear and non-linear effects of age, sex, education, cardiovascular risk factors, and presence of depressive, anxiety, and substance use disorders. Individuals from lower income households showed signs of advanced brain white matter aging with greater mean diffusivity, lower neurite density, lower myelination, and lower iron content. The association between household income and mean diffusivity was mediated by neurite density (B=0.084, p=0.003) and myelination (B=0.019, p=0.009); mean diffusivity partially mediated the association between household income and cognitive performance (B=0.017, p<0.05). Household income moderated the relation between white matter microstructure and cognitive performance, such that greater mean diffusivity, lower myelination, or lower neurite density was only associated with poorer cognitive performance among individuals from lower income households. Individuals from higher income households showed preserved cognitive performance even with greater mean diffusivity, lower myelination, or lower neurite density. These findings provide novel mechanistic insight into associations between socioeconomic conditions, brain anatomy, and cognitive performance in middle to late adulthood. Significance statement Pathways linking socioeconomic conditions, brain anatomy, and cognitive performance have rarely been investigated. Using multi-contrast imaging, we found that individuals from lower income households had markers of advanced brain white matter aging with lower neurite density, lower myelination, and lower iron content, alongside greater mean diffusivity. Greater mean diffusivity (reflecting myelin and neurite density) contributed to the association between household income and cognitive performance. Household income also buffered the observed white matter effects, such that greater mean diffusivity, lower index of myelin content, or lower neurite density was only associated with poorer cognitive performance among individuals from lower income households. These findings provide a detailed neurobiological understanding of socioeconomic differences in brain anatomy and associated cognitive performance.

QUIQI II package includes supporting material for the scientific article by Corbin et al. entitled ‘Statistical analyses of motion-corrupted MRI relaxometry data’. The complete support package for the QUIQI method includes: A copy of the original analysis code used to compile the results presented in the original scientific publication (doi: 10.5281/zenodo.7612032) A subset of the data used in the original publication for computation of the results. This data also includes a set of analysis results obtained by running the code described in 1. on the provided data. The combination of 1. and 2. allows users to replicate the computation of the provided analysis results. The material provided here only concerns part 2. of the QUIQI support package described above - subset of the data used in the original publication.

Consistent noise variance across data points (i.e. homoscedasticity) is required to ensure the validity of statistical analyses of MRI data conducted using linear regression methods. However, head motion leads to degradation of image quality, introducing noise heteroscedasticity into ordinary-least square analyses. The recently introduced QUIQI method restores noise homoscedasticity by means of weighted least square analyses in which the weights, specific for each dataset of an analysis, are computed from an index of motion-induced image quality degradation. QUIQI was first demonstrated in the context of brain maps of the MRI parameter R2*, which were computed from a single set of images with variable echo time. Here, we extend this framework to quantitative maps of the MRI parameters R1, R2*, and MTsat, which are computed from multiple sets of images. QUIQI allows for optimization of the noise model by using metrics quantifying heteroscedasticity and free energy. QUIQI restores homoscedasticity more effectively than insertion of an image quality index in the analysis design and yields higher sensitivity than simply removing the datasets most corrupted by head motion from the analysis. In sum, QUIQI provides an optimal approach to group-wise analyses of a range of quantitative MRI parameter maps that is robust to inherent homoscedasticity.

IntroductionIn this study, we applied multivariate methods to identify brain regions that have a critical role in shaping the connectivity patterns of networks associated with major psychiatric diagnoses, including schizophrenia (SCH), major depressive disorder (MDD) and bipolar disorder (BD) and healthy controls (HC). We used T1w images from 164 subjects: Schizophrenia (n = 17), bipolar disorder (n = 25), major depressive disorder (n = 68) and a healthy control group (n = 54).MethodsWe extracted regions of interest (ROIs) using a method based on the SHOOT algorithm of the SPM12 toolbox. We then performed multivariate structural covariance between the groups. For the regions identified as significant in t term of their covariance value, we calculated their eigencentrality as a measure of the influence of brain regions within the network. We applied a significance threshold of p = 0.001. Finally, we performed a cluster analysis to determine groups of regions that had similar eigencentrality profiles in different pairwise comparison networks in the observed groups.ResultsAs a result, we obtained 4 clusters with different brain regions that were diagnosis-specific. Cluster 1 showed the strongest discriminative values between SCH and HC and SCH and BD. Cluster 2 had the strongest discriminative value for the MDD patients, cluster 3 – for the BD patients. Cluster 4 seemed to contribute almost equally to the discrimination between the four groups.DiscussionOur results suggest that we can use the multivariate structural covariance method to identify specific regions that have higher predictive value for specific psychiatric diagnoses. In our research, we have identified brain signatures that suggest that degeneracy shapes brain networks in different ways both within and across major psychiatric disorders.

This chapter provides an overview of multivariate methods and several advanced demonstrations of how these methods can be applied to machine learning applications in mental health. Multivariate methods can help identify relationships between different variables (symptoms, behavior, brain anatomy, brain function, genetics, etc.) that may influence mental health. The methods provide valuable insights into how best to target interventions for a particular brain disorder. When using these methods, it is important to know the advantages, but also the limitations, of each method and how they can be applied in different contexts. In this chapter, we define and explain the concept of multivariate analyses from the theoretical basis to practical issues related to data preparation and interpretation of results. The methods described here include factor analysis, principal component analysis, path analysis, partial least squares, and linear multivariate methods. The chapter includes examples of how multivariate methods have been used, for example, to predict treatment outcomes or identify risk factors for mental disorders. We not only discuss the many challenges of using these methods in computational neuroscience research, such as the need for large and diverse datasets, but also introduce new approaches, such as guided model-based approaches and advanced AI-based approaches, to enrich these mostly data-driven methods and obtain better insight that integrates a priori information.

There is an ongoing debate about differential clinical outcome and associated adverse effects of deep brain stimulation (DBS) in Parkinson's disease (PD) targeting the subthalamic nucleus (STN) or the globus pallidus pars interna (GPi). Given that functional connectivity profiles suggest beneficial DBS effects within a common network, the empirical evidence about the underlying anatomical circuitry is still scarce. Therefore, we investigate the STN and GPi-associated structural covariance brain patterns in PD patients and healthy controls. We estimate GPi's and STN's whole-brain structural covariance from magnetic resonance imaging (MRI) in a normative mid- to old-age community-dwelling cohort (n = 1184) across maps of grey matter volume, magnetization transfer (MT) saturation, longitudinal relaxation rate (R1), effective transversal relaxation rate (R2*) and effective proton density (PD*). We compare these with the structural covariance estimates in patients with idiopathic PD (n = 32) followed by validation using a reduced size controls' cohort (n = 32). In the normative data set, we observed overlapping spatially distributed cortical and subcortical covariance patterns across maps confined to basal ganglia, thalamus, motor, and premotor cortical areas. Only the subcortical and midline motor cortical areas were confirmed in the reduced size cohort. These findings contrasted with the absence of structural covariance with cortical areas in the PD cohort. We interpret with caution the differential covariance maps of overlapping STN and GPi networks in patients with PD and healthy controls as correlates of motor network disruption. Our study provides face validity to the proposed extension of the currently existing structural covariance methods based on morphometry features to multiparameter MRI sensitive to brain tissue microstructure.