Abstract Background Major depressive disorder (MDD) is clinically heterogeneous, hindering identification of reproducible biomarkers. Using a semi-supervised machine learning approach, HYDRA, we previously identified two neuroanatomical dimensions from structural MRI in medication-free MDD from COORDINATE-MDD consortium. These dimensions (D1, D2) showed differential responses to selective serotonin reuptake inhibitor (SSRI) antidepressants and placebo. External replication in UK Biobank linked D2, characterized by widespread subtle neuroanatomical reductions, to an immuno-metabolic profile. Here, we examined whether these dimensions are detectable early in the course of illness. Methods We applied the pre-trained model to structural MRI data from the multisite PRONIA cohort, comprising individuals with recent-onset depression (ROD; n = 377; mean age 25.8 years, SD 6.0; 51.3% female) and healthy controls (n = 267; mean age 25.5 years, SD 6.4; 61.0% female). Participants were assigned to clusters (C1, C2) corresponding to the previously identified dimensions (D1, D2). Clusters were compared on clinical symptom profiles, peripheral inflammatory markers, and in a subset (n = 107), proteomic ageing indices. Results Two neuroanatomical clusters were identified in PRONIA. C1 (n = 265) showed higher negative symptom severity and elevated interleukin-2 levels. C2 (n = 140) was associated with higher residual proteomic age. Overall depressive symptom severity did not differ significantly between clusters. Conclusions Neuroanatomical dimensions of MDD are reproducible and detectable at illness onset. Associations with negative symptom severity, inflammatory signalling, and proteomic ageing suggest these dimensions capture biologically meaningful heterogeneity early in depression. These findings support a biologically informed framework for stratified treatment approaches in MDD.

Major depressive disorder (MDD) is common and disabling, yet reported brain structural differences vary across studies. Here we performed a large vertex-wise (point-by-point) meta-analysis of cortical thickness and surface area using harmonized magnetic resonance imaging processing across 64 cohorts from the Enhancing NeuroImaging Genetics through Meta-Analysis (ENIGMA) MDD and Depression Imaging Research Consortium (DIRECT) consortia (5,736 patients; 6,538 controls). We show significantly lower cortical thickness in patients with MDD in multiple brain regions, including the inferior parietal, lateral occipital, superior parietal, medial and lateral orbitofrontal, anterior and posterior cingulate, and precentral gyri, with cortical surface area showing no significant differences. Effects were most pronounced in adults with acute depression, whereas adolescents showed no significant case-control differences. Antidepressant medication use at scanning was associated with more extensive thinning, although effect sizes remained modest (mostly |Cohen's d| < 0.20). This high-resolution, globally generalizable map can support studies of mechanisms and help evaluate structural markers of the clinical course and treatment response.

Subjective well-being is a critical protective factor during childhood and adolescence, yet its neural mechanisms remain poorly characterized, in part because prior neuroimaging work has relied on unidimensional hedonic scales and pathological cohorts. We examined whole-brain neural correlates of multidimensional well-being in a typically developing Chinese cohort (N = 196, ages 6–18 years) using resting-state functional MRI (rs-fMRI), network-based statistics, and graph-theoretical analysis. Well-being was assessed with the PERMA-Profiler, in which higher scores indicate greater flourishing across five domains. Higher overall well-being, driven mainly by the Engagement dimension, was associated with reduced functional connectivity (FC) across large-scale networks. The strongest pattern involved functional decoupling between the default mode network (DMN) and task-positive networks and between the somatomotor and ventral attention networks. Complementary topological analyses linked higher engagement to lower global network integration. Edge-level FC associations survived stringent correction for multiple comparisons, and engagement-specific topological associations proved more robust than total well-being scores after accounting for nonlinear age effects. Rather than indicating dysfunction, this sparse connectivity profile may reflect a more economical, lower-density resting-state architecture in healthy development. These findings position engagement as a candidate bridging between positive psychological functioning and intrinsic brain network organization in youth, and demonstrate that multidimensional well-being frameworks can reveal domain-specific neural signatures invisible to unidimensional approaches.

Background:Major depressive disorder (MDD) is associated with altered brain structure and evidence of accelerated brain aging. However, previous studies have been limited by clinical samples with mixed medication status and multiple mood states, modest sample sizes, small percentage of MDD individuals older than 65 years of age, and/or reliance on summary-level data. Methods:Harmonized T1-weighted MRI from MDD (n = 645), all medication-free and in a current depressive episode, and matched healthy controls (n = 645), segmented into 145 regional volumes, from 11 sites in COORDINATE-MDD consortium. Brain age gap (BAG) was estimated using gradient boosting regression with nested cross-validation. Group differences in BAG (and age-corrected BAG [cBAG]) were examined across age strata. Regional contributions were evaluated using Shapley Additive exPlanations. Results:MDD was associated with significantly elevated cBAG compared with healthy controls (mean difference + 2.01 years). Age-stratified analyses showed no differences before mid-30s, with progressively larger gaps thereafter, reaching +6.85 years in MDD aged 55 and older. cBAG differed across neuroanatomical phenotypes associated with differential antidepressant response, cognitive impairment, increased adverse life events, increased self-harm and suicide attempts, and a pro-atherogenic metabolic profile. Key contributing regions included lateral and medial prefrontal regions, middle temporal gyrus, putamen, supplementary motor cortex, central operculum, and cerebellum. Conclusions:Accelerated structural brain aging in MDD is age-dependent and is most pronounced in a neuroanatomical phenotype associated with worse key clinical outcomes. The findings support neuroprogression models of MDD while demonstrating that cBAG is not a uniform feature of MDD and seem to be more strongly expressed in a specifically clinically vulnerable disease phenotype.

Functional magnetic resonance imaging (fMRI) allows real-time observation of brain activity through blood oxygen level-dependent (BOLD) signals and is extensively used in studies related to sex classification, age estimation, behavioral measurements prediction, and mental disorder diagnosis. However, the application of deep learning techniques to brain fMRI analysis is hindered by the small sample size of fMRI datasets. Transfer learning offers a solution to this problem, but most existing approaches are designed for large-scale 2D natural images. The heterogeneity between 4D fMRI data and 2D natural images makes direct model transfer infeasible. This study proposes a novel geometric mapping-based fMRI transfer learning method that enables transfer learning from 2D natural images to 4D fMRI brain images, bridging the transfer learning gap between fMRI data and natural images. The proposed Multi-scale Multi-domain Feature Aggregation (MMFA) module extracts effective aggregated features and reduces the dimensionality of fMRI data to 3D space. By treating the cerebral cortex as a folded Riemannian manifold in 3D space and mapping it into 2D space using surface geometric mapping, we make the transfer learning from 2D natural images to 4D brain images possible. Moreover, the topological relationships of the cerebral cortex are maintained with our method, and calculations are performed along the Riemannian manifold of the brain, effectively addressing signal interference problems. The experimental results based on the Human Connectome Project (HCP) dataset demonstrate the effectiveness of the proposed method. Our method achieved state-of-the-art performance in sex classification, age estimation, and behavioral measurement prediction tasks. Moreover, we propose a cascaded transfer learning approach for depression diagnosis, and proved its effectiveness on 23 depression datasets. In summary, the proposed fMRI transfer learning method, which accounts for the structural characteristics of the brain, is promising for applying transfer learning from natural images to brain fMRI images, significantly enhancing the performance in various fMRI analysis tasks.

Major depressive disorder (MDD) imposes significant global health burdens, yet its underlying neural mechanisms remain elusive. Traditional static functional metrics inadequately capture the brain’s dynamic nature, motivating the exploration of dynamic functional metrics to understand both the temporal and spatial reconfigurations of brain networks in MDD. Leveraging the Depression Imaging Research Consortium (DIRECT) dataset, this study conducted vertex-wise dynamic analyses in a large cohort of MDD patients (n = 1660) and healthy controls (n = 1341). We identified significant alterations in temporal stability across the brain, with MDD patients exhibiting increased stability in higher-order association areas (e.g., frontoparietal and default mode networks) and decreased stability in primary sensory-motor regions. Among the regions showing altered temporal stability, brain-symptom relationships were further explored. We identified a set of brain regions including the superior frontal gyrus, postcentral gyrus and superior insular sulcus, which were potentially involved in the common abnormal dFC network and associated with insomnia, feelings of guilt, and insight symptoms in MDD. By incorporating advanced vertex-wise dynamic functional analyses and a large sample size, this study provides insights into the neural mechanisms of MDD, emphasizing the value of dynamic approaches for identifying biomarkers. Future longitudinal and task-based studies are promising to elucidate causal relationships and refine personalized therapeutic interventions targeting specific dynamic dysfunctions in MDD.

As a prominent psychopathological process of major depressive disorder (MDD), rumination’s brain underpinnings remain unclear. Emerging studies have highlighted that the brain is organized along several macroscale gradients, which could serve as a powerful framework for better understanding how the functional connectome underlies rumination. In this study, we leveraged two datasets (Rum-Beijing and Rum-MDD) to characterize the gradient structure during an active ruminative state. Rum-Beijing consisted of 40 healthy controls (HC) who underwent 3 repetitive scans, while Rum-MDD consisted of 45 patients with major depressive disorder (MDD) and 46 HCs. We used a modified rumination state task (RST) to induce participants into a continuous, active rumination state and investigated the gradient profiles. Several global features of the gradient (range, explanation ratio, variance) and the regional differences of each gradient were also compared. In the Rum-MDD dataset, we further examined the interaction effect between the group (MDD vs. HC) and condition (rumination vs. distraction) regarding the gradient’s global and local metrics. Two gradients were identified: the primary-transmodal gradient and the visual-sensorimotor gradient. We found that the rumination state exhibited reduced gradient values in the default mode network (DMN) as compared to the distraction state. We identified replicable altered gradient values in the left dorsolateral prefrontal cortex (DLPFC), superior frontal gyrus (SFG), and posterior cingulate cortex (PCC). Relative to the distraction state, individuals’ gradient range exhibited a replicable and significant reduction along the primary–transmodal gradient, accompanied by a significant increase along the visual–sensorimotor gradient during rumination. Moreover, the MDD group showed a significantly higher primary–transmodal gradient range during the rumination state than HCs. In conclusion, the present study showed that rumination may correspond to a specific underlying functional gradient profile, which was altered in patients with MDD. These results shed new light on the neural mechanisms underlying rumination, highlighting a global functional coupling characteristic across the whole brain during active rumination.

Objective Although previous studies have reported structural brain alterations in major depressive disorder (MDD) patients at risk of suicide, no wide consensus has been reached. This study aimed to elucidate structural brain differences between MDD patients with and without suicide risk using data from the DIRECT Consortium, advancing our understanding of the neurophysiological mechanisms underlying suicide risk in MDD patients. Methods A total of 203 healthy controls (HCs), 208 MDD patients without suicide risk (MDD-NSR), and 376 MDD patients with suicide risk (MDD-SR) were included. T1-weighted MRI data were processed using DPABISurf to quantify cortical surface area, thickness, and cortical gray matter (CGM) volume. Results The MDD-SR and MDD-NSR groups demonstrated reduced surface area in the right orbitofrontal cortex (OFC), and reduced CGM volume in the left ACC and the right OFC compared with the HC group. The left ACC CGM volume decreased in a gradient, with the MDD-SR group showing significantly lower values than both the MDD-NSR and HC groups. While the SVM model showed limited differentiation between MDD-SR and MDD-NSR, the left ACC CGM volume was identified as the most critical feature in SHAP analysis. Conclusion MDD patients with suicide risk exhibited alterations in cortical surface area and CGM volume in the frontal lobe. The CGM volume reduction in the left ACC may serve as a potential biomarker for predicting suicide risk in MDD patients.

Major depressive disorder (MDD) has been increasingly characterized as a network dysconnectivity syndrome. Although single-subject morphological networks are advantageous in studying the brain connectome, extant research on MDD is limited by either small samples or a lack of integration of multi-feature across different morphological features. We used the largest structural MRI data from 1442 MDD patients and 1277 controls to construct individual-level cortical morphological networks based on cortical thickness (CT), cortical volume (CV), surface area (SA), and sulcal depth (SD). Group comparisons in interregional morphological connectivity (MC) and graph-theoretical nodal properties were performed. Furthermore, support vector machine (SVM) was applied to evaluate whether the network alterations could distinguish patients from controls. As a result, MDD patients presented widespread alterations in MC, with distinct alteration patterns observed across four morphological networks. Specifically, CT-based networks exhibited reduced MC primarily within and between higher-order networks involving the default mode and frontoparietal networks, whereas CV-based networks showed increased MC predominantly within the default mode network. By contrast, both SA- and SD-based networks demonstrated enhanced MC mainly within and between lower-order networks implicating the somatomotor and visual networks. Similar patterns of MC alterations were observed in first-episode, drug-naive MDD patients. Concurrently, nodal property analysis revealed increased betweenness centrality in multiple cortical regions in MDD. Moreover, SVM models based on the altered MC achieved moderate-to-good classification performance in distinguishing patients from controls. Overall, our findings of individual-level morphological network alterations in depressed patients may corroborate the dysconnectivity hypothesis of MDD and could further inform its more accurate diagnosis.

Magnetic resonance imaging (MRI) biomarkers have shown considerable potential in elucidating the neurobiological underpinnings of major depressive disorder (MDD). However, clinical translation of these biomarkers remains limited due to reliance on group-level analyses, which fail to capture the individual variability inherent in MDD. Precision psychiatry, which advocates for individualized approaches, offers a framework that could enhance the clinical utility of MRI biomarkers across multiple domains, including diagnostic classification, treatment response prediction, and individualized interventions. Despite this potential, current research applying MRI biomarkers to MDD within the framework of precision psychiatry remains fragmented, lacking an integrated clinical system that seamlessly combines these components. This review introduces the concept of a closed-loop clinical system, emphasizing the integration of diagnostic classification, treatment response prediction, and individualized interventions into a unified approach at the individual patient level. We summarize recent advances in these three clinical domains, highlight existing fragmentation, and discuss the challenges of achieving a cohesive system. Finally, we propose that the integration of MRI biomarkers into a closed-loop clinical system, as envisioned by precision psychiatry, holds great promise for the individualized management of MDD, improving clinical outcomes from diagnosis through recovery.

Aberrant functional connectivity (FC) between brain networks has been indicated closely associated with bipolar disorder (BD). However, the previous findings of specific brain network connectivity patterns have been inconsistent, and the clinical utility of FCs for predicting treatment outcomes in bipolar depression was underexplored. To identify robust neuro-biomarkers of bipolar depression, a connectome-based analysis was conducted on resting-state functional MRI (rs-fMRI) data of 580 bipolar depression patients and 116 healthy controls (HCs). A subsample of 148 patients underwent a 4-week quetiapine treatment with post-treatment clinical assessment. Adopting machine learning, a predictive model based on pre-treatment brain connectome was then constructed to predict treatment response and identify the efficacy-specific networks. Distinct brain network connectivity patterns were observed in bipolar depression compared to HCs. Elevated intra-network connectivity was identified within the default mode network (DMN), sensorimotor network (SMN), and subcortical network (SC); and as to the inter-network connectivity, increased FCs were between the DMN, SMN and frontoparietal (FPN), ventral attention network (VAN), and decreased FCs were between the SC and cortical networks, especially the DMN and FPN. And the global network topology analyses revealed decreased global efficiency and increased characteristic path length in BD compared to HC. Further, the support vector regression model successfully predicted the efficacy of quetiapine treatment, as indicated by a high correspondence between predicted and actual HAMD reduction ratio values (r(df=147)=0.4493, p = 2*10-4). The identified efficacy-specific networks primarily encompassed FCs between the SMN and SC, and between the FPN, DMN, and VAN. These identified networks further predicted treatment response with r = 0.3940 in the subsequent validation with an independent cohort (n = 43). These findings presented the characteristic aberrant patterns of brain network connectome in bipolar depression and demonstrated the predictive potential of pre-treatment network connectome for quetiapine response. Promisingly, the identified connectivity networks may serve as functional targets for future precise treatments for bipolar depression.

Background Rumination is a pivotal psychopathological process in major depressive disorder (MDD). The neurotrophic hypothesis suggests that glial cell line-derived neurotrophic factor (GDNF) might play a role in brain dysfunction and clinical symptoms of MDD. However, the relationship remains unclear. Methods Thirty-three individuals with MDD and 33 healthy controls (HCs) underwent functional magnetic resonance imaging (fMRI) while performing a rumination state task designed to induce sustained, active rumination. The Ruminative Response Scale (RRS) was administered to assess individual rumination tendency. Brain activity within the default mode network (DMN) subsystems during rumination was characterized using both fractional amplitude of low-frequency fluctuations (fALFF) and functional connectivity (FC) analyses. Serum levels of GDNF and inflammatory markers [interleukin (IL)-6, IL-8, and C-reactive protein] were quantified in all participants. We then examined the relationships between regional brain activity (fALFF values), GDNF levels, and rumination severity (RRS scores) in the MDD group. Results Compared to HCs, MDD patients exhibited significantly reduced serum levels of both GDNF (t = −3.204, P = 0.002) and IL-8 (t = −3.239, P = 0.002). Significant interaction effects were observed in fALFF within both the dorsal medial prefrontal cortex (DMPFC; F = 25.075, P < 0.001) and medial temporal lobe (MTL; F = 28.753, P < 0.001) subsystems of the DMN. Mediation analysis revealed that the relationship between GDNF levels and brooding rumination in MDD patients was mediated by neural activity within the DMPFC subsystem. Conclusions In MDD patients, GDNF levels were associated with neural activity within the DMPFC subsystem of the DMN, which statistically mediated the link to rumination severity.

Background:Rumination is a pivotal psychopathological process in major depressive disorder (MDD). The neurotrophic hypothesis suggests that glial cell line-derived neurotrophic factor (GDNF) might play a role in brain dysfunction and clinical symptoms of MDD. However, the relationship remains unclear. Methods:Thirty-three individuals with MDD and 33 healthy controls (HCs) underwent functional magnetic resonance imaging (fMRI) while performing a rumination state task designed to induce sustained, active rumination. The Ruminative Response Scale (RRS) was administered to assess individual rumination tendency. Brain activity within the default mode network (DMN) subsystems during rumination was characterized using both fractional amplitude of low-frequency fluctuations (fALFF) and functional connectivity (FC) analyses. Serum levels of GDNF and inflammatory markers [interleukin (IL)-6, IL-8, and C-reactive protein] were quantified in all participants. We then examined the relationships between regional brain activity (fALFF values), GDNF levels, and rumination severity (RRS scores) in the MDD group. Results:Compared to HCs, MDD patients exhibited significantly reduced serum levels of both GDNF (t = -3.204, P = 0.002) and IL-8 (t = -3.239, P = 0.002). Significant interaction effects were observed in fALFF within both the dorsal medial prefrontal cortex (DMPFC; F = 25.075, P < 0.001) and medial temporal lobe (MTL; F = 28.753, P < 0.001) subsystems of the DMN. Mediation analysis revealed that the relationship between GDNF levels and brooding rumination in MDD patients was mediated by neural activity within the DMPFC subsystem. Conclusions:In MDD patients, GDNF levels were associated with neural activity within the DMPFC subsystem of the DMN, which statistically mediated the link to rumination severity.

BACKGROUND:Cortical morphological alterations are evident in major depressive disorder (MDD), yet the underlying neurobiological processes that contribute to their characteristic spatial pattern remain unclear. METHODS:Large-scale, multisite structural magnetic resonance imaging data from a homogeneous Chinese cohort of 1442 patients with MDD and 1277 healthy control participants were used to calculate cortical morphological measures, which were compared between groups to determine cortical morphological alterations in MDD. A connectome constraint model was then used to examine whether the structural connectome shapes MDD-related cortical morphological alterations, followed by performance of a network diffusion model to identify the epicenters. RESULTS:Group comparisons demonstrated a broadly distributed cortical thickness (CT) reduction in MDD, with the prefrontal cortex affected more prominently. Based on the normative structural connectome, we derived the estimated CT alteration of each brain node according to its connected neighbors and found a strong spatial correlation between the empirical and estimated CT alterations, indicating structural connectome constraint on cortical atrophy in MDD. Concurrently, we identified the left lateral prefrontal cortex as the putative epicenter of cortical atrophy. Moreover, analyses across first-episode, early-stage, and chronic MDD subgroups revealed reduced connectome constraint with increasing illness duration. Additionally, our results were robust against several methodological variations and were largely reproducible in the cross-ethnic ENIGMA (Enhancing Neuro Imaging Genetics through Meta Analysis) cohort of 1902 patients with MDD and 7658 control participants. CONCLUSIONS:These findings represent a substantial advance in our understanding of the network-based spread of cortical atrophy in MDD and highlight the prospect of the left prefrontal cortex as a key target for early interventions.

Hemispheric lateralization, recognized as a pivotal feature in both the structural and functional organization of the human brain, may undergo alterations in specific psychiatric disorders. However, the time-varying patterns of hemispheric lateralization in individuals with major depressive disorder (MDD) and the relationship between these patterns and gene expression profiles remain largely unexplored thus far. Using a large multi-site resting-state functional magnetic resonance imaging (rs-fMRI) data encompassing 2611 participants (1660 MDD patients and 1341 healthy controls), we examined MDD-related abnormalities in dynamic laterality and its association with clinical symptoms, meta-analytic cognitive functions, and neurotransmitter receptor profiles, respectively. And the biological basis behind these changes was investigated through gene enrichment analysis and cell-specific analysis. Here we found revealed pronounced fluctuations in lateralization primarily in the regions in default mode network, attention network and control network in MDD patients when compared to healthy controls. In addition, these fluctuations exhibited significant correlations with higher-order cognition terms and the distributions of disease related neurotransmitters. Further, through gene enrichment and cell-specific analysis, we identified a molecular genetic basis for these changes, highlighting synaptic function-related genes and neuronal cells. Collectively, these results demonstrated robust altered brain lateralization patterns in MDD and its molecular genetic basis, providing new clues to understand the pathophysiology of MDD.

Background Suicidal ideation (SI) is very common in patients with major depressive disorder (MDD). However, its neural mechanisms remain unclear. The anterior cingulate cortex (ACC) region may be associated with SI in MDD patients. This study aimed to elucidate the neural mechanisms of SI in MDD patients by analyzing changes in gray matter volume (GMV) in brain structures in the ACC region, which has not been adequately studied to date. Methods According to the REST-meta-MDD project, this study subjects consisted of 235 healthy controls and 246 MDD patients, including 123 MDD patients with and 123 without SI, and their structural magnetic resonance imaging data were analyzed. The 17-item Hamilton Depression Rating Scale (HAMD) was used to assess depressive symptoms. Correlation analysis and logistic regression analysis were used to determine whether there was a correlation between GMV of ACC and SI in MDD patients. Results MDD patients with SI had higher HAMD scores and greater GMV in bilateral ACC compared to MDD patients without SI (all p < 0.001). GMV of bilateral ACC was positively correlated with SI in MDD patients and entered the regression equation in the subsequent logistic regression analysis. Conclusions Our findings suggest that GMV of ACC may be associated with SI in patients with MDD and is a sensitive biomarker of SI.

Social skills training (SST) has demonstrated efficacy in improving social deficits in individuals with autism spectrum disorder (ASD), but the underlying neural mechanisms remain unclear. This study investigated alterations in whole-brain white matter network topology after SST in ASD individuals and explored potential correlation with improvements in social interaction deficits. 38 ASD patients aged 12 - 30 years were recruited, including 19 who completed magnetic resonance imaging (MRI) scans and social responsiveness scale (SRS) assessments at both baseline and the endpoint of a 14-week SST (training group) and 19 age-, sex-, and IQ-matched patients who underwent MRI scans and SRS assessment at the same time points but did not receive SST (control group). White matter connectivity matrices were constructed using diffusion tensor imaging (DTI), and graph theory analysis was used to assess global and nodal network properties. Paired t-tests and independent-samples t-tests were used for within- and between-group comparisons, respectively. Pearson's partial correlation was used to examine associations between network changes and SRS scores changes. After SST, four edges showed significant changes in white matter connectivity (FDR-corrected), with three increased and one decreased in the training group. Changes in nodal betweenness were also observed. While SRS scores significantly decreased in the training group, no significant correlations were found between neuroimaging changes and behavioral improvements, possibly due to the limited sample size. These findings suggest that SST may reshape white matter network, offering insights into its neural mechanisms and informing novel ASD intervention strategies.

Timely intervention for Alzheimer's disease (AD) requires early detection. The development of immunotherapies targeting amyloid-beta and tau underscores the need for accessible, time-efficient biomarkers for early diagnosis. Here, we directly applied our previously developed MRI-based deep learning model for AD to the large Chinese SILCODE cohort (722 participants, 1,105 brain MRI scans). The model - initially trained on North American data - demonstrated robust cross-ethnic generalization, without any retraining or fine-tuning, achieving an AUC of 91.3% in AD classification with a sensitivity of 95.2%. It successfully identified 86.7% of individuals at risk of AD progression more than 5 years in advance. Individuals identified as high-risk exhibited significantly shorter median progression times. By integrating an interpretable deep learning brain risk map approach, we identified AD brain subtypes, including an MCI subtype associated with rapid cognitive decline. The model's risk scores showed significant correlations with cognitive measures and plasma biomarkers, such as tau proteins and neurofilament light chain (NfL). These findings underscore the exceptional generalizability and clinical utility of MRI-based deep learning models, especially in large and diverse populations, offering valuable tools for early therapeutic intervention. The model has been made open-source and deployed to a free online website for AD risk prediction, to assist in early screening and intervention.

BACKGROUND:The Hamilton Rating Scale for Depression (HRSD) and the Montgomery-Åsberg Depression Rating Scale (MADRS) are the two most common clinician-rated scales to quantify depression symptom change in repetitive transcranial magnetic stimulation (rTMS) trials. However, it is unclear how the values of one scale translate to the other. Being able to translate scores between these scales could allow for aggregating rTMS clinical trial data. METHODS:Clinical data from two randomised rTMS clinical trials (FOURD and CARTBIND, total N=380) were pooled. We used five crosswalk models: (1) a pharmacotherapy equipercentile model, (2) an rTMS equipercentile model, (3) a linear regression model, (4) a random forest (RF) regression model and (5) a support vector regression (SVR) model. Model performance was benchmarked using the root mean square error (RMSE). RESULTS:The linear regression model demonstrated the best performance (RMSE: 2.66-4.82), though the SVR model's performance was slightly worse but comparable (RMSE: 2.69-5.32). The RF regression model generally performed worst (RMSE: 2.70-5.20). The rTMS equipercentile model's performance was intermediate (RMSE: 2.69-5.32) in the primary analysis but achieved superior performance and demonstrated less bias in the additional analysis. INTERPRETATION:MADRS and HRSD scores from rTMS trials can be accurately converted between each other. The optimal model was the newly developed equipercentile model, though the results of the SVR model were promising. Nevertheless, independent external replication is required to demonstrate the external validity of these findings. TRIAL REGISTRATION NUMBER:FOURD: NCT02998580; CARTBIND: NCT02729792.