BACKGROUND:The hippocampus influences the outcomes of amnestic mild cognitive impairment (aMCI) and undergoes different changes during the cognitive decline or recovery of aMCI compared to elderly individuals with normal cognition, which may reveal disease-dependent neurodegeneration or plasticity. We first aimed to investigate the hippocampal changes associated with cognitive changes in aMCI using a combined case-control study design. METHODS:In total, 50 aMCI individuals and 50 healthy controls (HCs) were recruited in Shenyang, China, and separately randomized into training and control groups: aMCI training group, aMCI no training group, HC training group, and HC no training group. The aMCI and HC training groups received computerized cognitive training (CCT) thrice weekly for 12 weeks. Cognitive assessments and MRI data were collected at baseline and follow-up. RESULTS:The primary outcome was significant CCT×diagnosis interaction effect on the change in cognitive performance as measured by clock drawing test (CDT) scores (F = 4.322, P = 0.041); this interaction was driven by CCT specifically in aMCI (F = 4.465, P = 0.038). Significant CCT×diagnosis interaction effects of right-hippocampal FC changes were observed in the bilateral precuneus/cuneus (Pvoxel<0.05) driven by CCT in aMCI (F = 5.429, P = 0.023), and in the left superior temporal gyrus/middle temporal gyrus (STG/MTG, Pvoxel<0.05), driven by CCT of only in HCs (F = 6.587, P = 0.013). A significant interaction effect of left-hippocampal FC changes were observed in the right triangular part of the inferior frontal gyrus (IFGtriang, Pvoxel<0.05), driven by CCT in aMCI and HCs (F = 6.550, P = 0.013; F = 7.097, P = 0.010). No significant interaction effect on the change in hippocampal GMV was noted (P > 0.05). CONCLUSION:CCT can improve the visuospatial ability of aMCI, which is reflected by the CDT scores. CCT can alter hippocampal FC in the bilateral precuneus/cuneus, the right IFGtriang, and the left STG/MTG. The hippocampal GMV is difficult to change in both HCs and aMCI during the cognitive decline. REGISTRATION NUMBER:ChiCTR1900026849. DATE OF REGISTRATION:24 October 2019 NAME OF TRIAL REGISTRY: Chinese Clinical Trial Registry (ChiCTR).

Tardive dyskinesia (TD) may reflect an intrinsic neurobiological vulnerability associated with schizophrenia, yet metabolic alterations in the caudate nucleu, a key brain region regulating involuntary movements-remain poorly characterized by proton magnetic resonance spectroscopy (1 H-MRS). We investigated the relationship between caudate nucleus metabolite concentrations and TD symptoms using 1 H-MRS. We recruited 117 patients with schizophrenia, including 67 patients with TD and 50 patients without TD (NT). We also recruited 41 healthy controls (HCs). Absolute metabolite concentrations of N-acetylaspartate plus N-acetyl-aspartyl-glutamate (tNAA), creatine (Cr) and glutamine plus glutamate (Glx) in the caudate nucleus were quantified using a 3.0-T MRI scanner with water signal referencing. Clinical symptoms were assessed using the Positive and Negative Syndrome Scale (PANSS) and the Abnormal Involuntary Movement Scale (AIMS). Among-group differences were analyzed using analysis of covariance with sex, age and years of education as covariates, followed by Bonferroni correction for multiple comparisons. After correction, the adjusted p-value was 0.017. Regression analysis was performed on metabolic indexes in the TD group with disease duration as a covariate. No significant differences in age, sex, or education level were observed among the groups. Additionally, no significant differences in PANSS scores or antipsychotic medication dosage were observed between the TD and NT groups. The TD group exhibited a significantly longer disease duration than the NT group (p < 0.05). Absolute tNAA levels—a marker of neuronal integrity—were significantly lower in the TD group than in the NT group (p < 0.05), while Cr (involved in energy metabolism) and Glx (involved in excitatory neurotransmission) levels did not significantly differ. These findings suggest reduced neuronal viability in the caudate nucleus in TD, though the clinical implications warrant further investigation.

Abstract Large language model (LLM) systems can now generate complete research manuscripts, yet their reliability in clinical medicine — where citation accuracy and reporting standards carry direct consequences — has not been systematically assessed. We introduce MedResearchBench, a benchmark of three clinical epidemiology tasks built on NHANES data, and use it to evaluate six AI research systems across six quality dimensions. Evaluation combines programmatic citation verification, rule-based reporting compliance checks, and multi-model LLM judging, providing a more discriminative assessment than conventional single-judge approaches. Citation integrity emerged as the decisive quality dimension. Hallucination rates ranged from 2.9% to 36.8% across systems, and a hard-rule threshold on per-task citation scores capped four of six systems’ total scores at the penalty ceiling. Adding a multi-agent citation verification and repair pipeline to the best-performing system improved its citation integrity score from 40.0 to 90.9 and raised the weighted total from 68.9 to 81.8. Strikingly, a single-model evaluation ranked this system last (55.5), while our three-tier framework ranked it first (81.8) —a complete reversal that exposes the limitations of subjective LLM-only evaluation. These results suggest that programmatic citation verification should be a core metric in future evaluations of AI scientific writing systems, and that multi-agent quality assurance can bridge the gap between fluent text generation and trustworthy scholarship.

BACKGROUND:Interventions addressing emotion-related school attendance challenges in adolescents with depression often face limited efficacy, partly due to parent-adolescent perceptual biases. The effectiveness of dual-track family interventions remains underexplored. This study evaluated the preliminary clinical effectiveness of a dual-track parent-adolescent group cognitive behavioral therapy (GCBT) program and explored parent-adolescent perceptual bias as a potential process marker. METHODS:This quasi-experimental study included 71 adolescents with depression and school refusal. Participants were allocated to an 8-week dual-track GCBT (n = 39; adolescent offline and parent online sessions) or treatment-as-usual (TAU; n = 32). The primary outcome was family-reported school-return status at week 8. Secondary outcomes included cross-informant clinical symptoms. Exploratory process analyses involved longitudinal trajectories of parent-adolescent emotional perceptual biases within the GCBT group. RESULTS:At week 8, the proportion of adolescents attending school increased from 16.7% to 66.7% in the GCBT group, compared with 46.9-50.0% in the TAU group (GEE group × time interaction: β = 2.06, P = .001). For secondary outcomes, parent-reported SDQ total difficulties and emotional symptoms showed significantly greater improvements after FDR correction (adjusted P = .002 and.003, respectively), whereas adolescent-reported PHQ-9 and GAD-7 showed non-significant improvement trends after FDR correction (adjusted P = .085 and.084, respectively). Exploratory within-intervention process analyses showed a reduction in parental positive bias over the 8-week program, with parent-adolescent perceptual bias decreasing at an average rate of 3.2 points per week (P = .02). CONCLUSIONS:The scalable dual-track GCBT model demonstrated preliminary effectiveness improving family-reported school-return status among adolescents with depression and school refusal. Exploratory within-intervention process analyses suggested that reductions in parental cognitive misjudgments and parent-adolescent perceptual discrepancies may represent potential therapeutic process markers, offering a practical approach for real-world clinical settings.

Generative Adversarial Networks, a popular deep learning method, have achieved excellent performance in both classification and prediction tasks. However, there have been relatively few applications of generative adversarial networks to EEG data. To study the effect of high-order brain functional networks on schizophrenia patients, a high-order graph attention generative adversarial network prediction model is proposed, and the generator of the model utilizes graph attention networks and long short-term memory networks to capture the high-order topological features of persistence images for early diagnosis and prediction of schizophrenia patients. The research results on the five frequency bands of schizophrenia show that the proposed prediction model performs best in the Theta frequency band, with AUC and MAP values reaching 93.5% and 93.0%, respectively, and an average accuracy of 91.5%, both of which are superior to the selected comparison methods. Moreover, the image quality coefficient is used to quantify the realism and clarity of the images generated by the model. the image quality coefficients of schizophrenia patients were significantly correlated with the PANSS total scores in the Gamma and Theta bands, which provided a new idea for generative adversarial networks in the prediction of schizophrenia high-order topological features.

BackgroundAdolescent depression diagnosis currently relies primarily on subjective self-report questionnaires, with a notable lack of objective neurobiological biomarkers. This study aimed to compare the diagnostic performance of traditional psychological scales with a newly developed Multi-view Adaptive Graph Convolutional Network (MVA-GCN) based on resting-state electroencephalography (EEG), and to exploratorily examine the associations between MVA-GCN-derived brain network features and psychological resilience in adolescents.MethodsResting-state EEG data were collected from 44 adolescents with major depressive disorder (MDD) and 30 healthy controls (HCs). The MVA-GCN model integrated three parallel connectivity views—phase-locking value (PLV), Pearson correlation coefficient (PCC), and phase lag index (PLI)—via an adaptive fusion mechanism. We compared the classification performance of questionnaire-based machine learning models against the MVA-GCN, and further examined the correlations between model-highlighted network features and clinical symptom measures, including depression, anxiety, loneliness, rumination, and resilience, with age and sex included as covariates.ResultsQuestionnaire-based machine learning models achieved a mean classification accuracy of 86.43%, whereas the MVA-GCN attained 99.84% accuracy in the present dataset. Occipital and fronto-central regions contributed most to the model’s predictions. Increased gamma-band functional connectivity and reduced alpha-band power were identified as potential electrophysiological correlates of group differences. Exploratory correlational analyses at the nominal level (p < 0.05, uncorrected) suggested potential group-specific patterns in brain–resilience associations. A post-hoc deviation index analysis showed that, within the MDD group, greater deviation from the healthy reference brain–resilience association pattern was significantly correlated with more severe rumination, self-rated depression, clinician-rated depression severity, loneliness, and self-rated anxiety (all FDR q < 0.05).ConclusionIn this exploratory sample, the MVA-GCN demonstrated promising proof-of-concept discriminative capability for adolescent depression compared with traditional self-report scales. However, the brain–resilience association findings were not statistically significant after correction for multiple comparisons and should be interpreted with caution. These results suggest that MVA-GCN-derived network features warrant further investigation in larger, independent cohorts, but do not yet support their use as a clinically validated diagnostic tool.

White matter abnormalities are important for understanding schizophrenia and related behaviors. However, the relationship between a history of suicide attempts and alterations in white matter microstructure among individuals with schizophrenia, and its impact on neurocognition, remains unclear. This study scanned 283 individuals diagnosed with schizophrenia and 189 healthy controls. Fractional anisotropy (FA) derived from diffusion tensor imaging is used to assess white matter microstructure. Neurocognitive performance was assessed using the MATRICS Consensus Cognitive Battery. Compared to healthy controls, schizophrenia patients with a history of suicide attempts showed widespread reductions in white matter FA across multiple brain regions (FDR-corrected p < 0.05). In uncorrected analyses, patients with a history of suicide attempts showed a significant lower FA in the external capsule (EC) (p = 0.033, Cohen’s d = -0.26) compared to patients without such history; two-week suicidal ideation was also associated with lower FA in the inferior fronto-occipital fasciculus (p = 0.018, d = -0.29). After FDR correction, none remained significant (all adjusted p > 0.05). A significant interaction was observed between suicide‑attempt history and EC FA in relation to the social cognition T‑score (p for interaction = 0.022). Contrary to prior findings in mood disorders, we did not observe robust white matter microstructural alterations associated with suicide attempts in schizophrenia. EC FA was significantly associated with social cognition performance only in patients without a history of suicide attempts. Schizophrenia patients with suicide attempts show widespread FA reductions versus healthy controls. EC FA is associated with social cognition only in non‑attempters, with a significant interaction by suicide history. No FA differences between attempters and non-attempters survive FDR correction.

Sleep perception impairment (SPI) characterized by subjective-objective discrepancies in sleep, is common among patients with depression. Its neurophysiological mechanisms remain unclear. This study investigated associations between polysomnography (PSG)-derived sleep macro- and micro-architecture features and SPI in depressed patients. We enrolled 63 adults (aged 18-65) with DSM-V major depressive disorder. The participants were divided into two groups: the SPI (n = 26) and non-SPI (n = 37). All underwent overnight PSG and completed clinical assessments. We analyzed sleep macro-architecture and EEG micro-architecture, including sleep temporal entropy (STE), reflecting fragmentation of sleep-stage transitions. Logistic and linear regression models assessed predictors of SPI, adjusting for demographics, clinical, and sleep-related covariates. Despite comparable objective sleep duration, SPI patients significantly underestimated their sleep duration based on their post-PSG subjective sleep time estimates, reported poorer subjective sleep quality, exhibited lower EEG total power (median 9.7 vs. 12.5 kµV²; p = 0.003), decreased interhemispheric EEG symmetry (0.50 vs. 0.51; p = 0.02), and elevated high-frequency relative to slow-wave EEG activity. Higher EEG total power (OR = 0.35 per 1000 µV² increase) and greater EEG symmetry (OR = 0.47 per 0.01 increase) independently predicted reduced odds of SPI in adjusted models. EEG-derived biomarkers (spectral power, symmetry, entropy) may differentiate sleep perception phenotypes in depression, offering potential targets for tailored clinical interventions.

High-order brain networks have emerged as a critical focus in neuroimaging and electrophysiological research due to their unique advantages in analyzing abnormal functional connectivity in psychiatric disorders. However, most existing EEG-based topological analysis methods rely on static inter-channel correlations to construct brain networks, which fail to capture the temporal dynamics and transient multichannel synchronization inherent in EEG signals. This limitation hinders the effective characterization of interaction patterns within brain networks. To address this issue, we propose a topological feature extraction framework based on temporal sampling and persistent homology, which constructs high-dimensional point clouds along the time axis to capture multi-scale topological structures. We further develop a vectorization module, PHEEG-Net, that automatically embeds persistence diagrams into high-dimensional learnable representations and integrates them with time- and frequency-domain features for classification tasks. In experiments conducted on the Theta band of two schizophrenia EEG datasets, PHEEG-Net achieves classification accuracies of 94.91% and 88.10%, significantly outperforming several existing topological vectorization approaches. These results validate its superior discriminative power and generalization capability. This work offers a novel perspective on the application of topological learning in EEG analysis and the auxiliary diagnosis of psychiatric disorders.

OBJECTIVE:Motor imagery EEG (MI-EEG) decoding remains challenging due to low signal-to-noise ratios and pronounced inter-subject variability. Although end-to-end deep models reduce reliance on manual feature engineering, many existing architectures may introduce temporal leakage through non-causal operations and often rely on fixed spatial topologies that cannot accommodate subject- and trial-specific connectivity patterns. APPROACH:We propose MAGCANet, which integrates five core components: (i) a Multiscale Causal Convolution Module (MCCM) for hierarchical temporal encoding under explicit causal constraints, (ii) a Temporal Convolution Module (TCM) to capture complex temporal dynamics, (iii) an Adaptive Graph Convolution Module (AGCM) for sample-specific topology learning in latent space, (iv) a Multi-Head Self-Attention Module (MHSAM) for global feature aggregation, and (v) a Classification Block for final decision making. Together, these components enforce temporal causality, adapt spatial interactions to individual dynamics, and produce discriminative representations robust to inter-subject variability. RESULTS:On the BCI Competition IV-2a and IV-2b datasets, MAGCANet achieves strong single-subject accuracies of 88.58% and 91.13%, respectively. Under Leave-One-Subject-Out (LOSO) evaluation, the model maintains accuracies of 70.49% and 79.49%, demonstrating competitive and stable cross-subject generalization. MAGCANet is highly lightweight, with only 0.0194M parameters, and achieves low inference latency (2.23 ms). Qualitative analyses, including feature clustering and channel occlusion, further highlight the model's interpretability and its ability to capture relevant EEG patterns. SIGNIFICANCE:MAGCANet provides a robust and interpretable solution for MI-EEG decoding, balancing high precision with computational efficiency, and offering a reliable method for real-time BCI applications.

The human cortical functional hierarchy, spanning from primary sensorimotor to transmodal association regions, represents a fundamental principle of brain organisation. Here, we show lifespan changes in the sensorimotor-association (S-A) gradient in the cortical functional hierarchy using multimodal neuroimaging data from 33,247 participants aged 32 postmenstrual weeks to 80 years. We identify three critical neurodevelopmental milestones: initiation (third trimester to perinatal period), establishment (infancy to early childhood), and expansion-stabilisation (late childhood to adulthood). Pronounced gradient changes are predominantly observed during the first decade, with continued refinement extending into mid-adulthood. Spatiotemporally heterogeneous growth patterns in functional gradients align with evolutionary hierarchies, segregation-integration dynamics, structural maturation, and cognitive spectrum development, proceeding along a dominant S-A growth axis. These findings establish a unified neurodevelopmental framework that links connectome gradient dynamics to multifaceted functional and structural properties, advancing our understanding of cortical hierarchy maturation across the lifespan.

First-episode schizophrenia (FES) presents with substantial heterogeneity in cognitive performance, yet this variability is frequently overlooked in studies investigating disease pathology and treatment. This study aimed to quantify cognitive heterogeneity across domains and individuals, and to identify distinct cognitive subtypes in FES. We assessed cognitive performance using the MATRICS Consensus Cognitive Battery (MCCB) and clinical symptoms with the Positive and Negative Syndrome Scale (PANSS) in 271 patients with FES and 133 healthy controls (HC). Domain-specific impairment was quantified using Hedges' g, whereas quantile-based analyses were performed to characterize distributional heterogeneity in overall cognitive performance. Gaussian mixture clustering was subsequently applied to identify latent subtypes within FES group. Compared with HC, patients with FES demonstrated significant impairments across all assessed cognitive measures. However, the magnitude of impairment varied across measures, with Attention/Vigilance as the domain with the greatest impairment. Quantile-based analyses demonstrated that the FES-HC difference was significantly greater at the lower than at the upper end of the cognitive distribution, indicating that cognitive impairment was not uniformly distributed across patients. Clustering analysis identified two cognitive subtypes: one subgroup with severe global cognitive deficits and another with relatively preserved cognitive functioning. Crucially, these subtypes also differed significantly in their clinical symptom profiles. In conclusion, specific cognitive impairments may reflect the underlying pathology of early schizophrenia, and individuals can be reliably stratified into distinct cognitive phenotypes. These findings provide critical clinical insights that can directly inform targeted, stratified strategies in early psychosis from a cognitive perspective.

Background The co-occurrence of depressive and hypomanic symptoms complicates mood disorder assessment and treatment. However, the symptom pathways linking these domains remain poorly understood. This study examined the network structure of depressive and hypomanic symptoms to identify key bridge symptoms. Methods Cross-sectional data from 6,913 psychiatric outpatients were analyzed using the Beck Depression Inventory-II (BDI-II) and Hypomania Checklist-32 (HCL-32). A Gaussian Graphical Model estimated symptom associations, and bridge centrality indices identified symptoms linking depressive and hypomanic communities. A Network Comparison Test evaluated differences between a Depressive Symptoms (DS) group and a Hypomanic-like Symptoms (HyS) group. Results Two distinct but interconnected symptom communities were identified. Irritability (HCL26), suicidal thoughts (BDI-II9), and impatience (HCL25) showed the highest bridge strength, indicating strong cross-domain connectivity. Social activity (HCL15) and loss of interest (BDI-II12) had the highest bridge betweenness, suggesting important roles in linking behavioral activation and depressive symptoms. Although overall network structure did not differ significantly between groups, the DS group showed higher global strength, indicating stronger overall symptom interdependence. Conclusions Irritability, social over-activation, and related symptoms occupy central positions connecting depressive and hypomanic symptom communities. These bridge symptoms may represent transdiagnostic indicators of affective symptom connectivity and provide potential directions for future longitudinal studies examining symptom evolution and clinical relevance.

Acute stress is a primary risk factor for mental disorders. While prior virtual reality (VR) studies have utilized objective physiological measures to address biological heterogeneity, the specific temporal evolution and cross-system coordination of these responses remain less explored. To address this gap, we developed a multi-scenario virtual reality (VR) stress paradigm, and analyzed personality modulation of stress response dynamics in 44 healthy adults via high-density monitoring (continuous heart rate/heart rate variability/galvanic skin response; 9-timepoint cortisol sampling). The VR paradigm effectively elicited multi-system stress responses that are similar to those observed in the Trier Social Stress Test. Personality traits significantly modulated temporal trajectories (15/28 time × trait interactions p adj < 0.10) but poorly predicted peak intensity (3/36 regressions significant), suggesting personality primarily influences "how responses evolve" rather than "how strong." Psychoticism and trait anxiety predicted autonomic desynchronization, particularly sympathetic-parasympathetic decoupling during recovery (β = 0.339 h, p adj = 0.031). These findings advance a paradigm shift from "reactivity" to "regulatory capacity" in stress research, providing a scalable digital tool for individualized stress susceptibility assessment.

Abstract Objective Childhood poverty is a high-risk context that involves diverse adversities, making it difficult to understand how poverty confers later psychopathology risk and why some children remain resilient despite growing up in poverty. To address this heterogeneity, we quantified adversity-linked vulnerability as adversity-psychopathology coupling and tested whether childhood poverty amplifies this coupling and whether multilevel inhibitory-control profiles stratify vulnerability and resilience within poverty-exposed youth. Methods We analyzed 10,112 youth (48.4% female; mean age = 9.92 years) from the Adolescent Brain Cognitive Development Study, linking baseline cumulative early-life adversity (ELA) to later behavioral problems across 4 waves. In the stop-signal task fMRI subsample of 7,401 youth, semi-supervised clustering of inhibitory-control activation identified neurofunctional subtypes within poverty-exposed youth. We also tested temperamental inhibitory control as an additional moderator. Results Childhood poverty amplified the association between cumulative ELA and behavioral problems at baseline (Δβ = 0.088; P < .001) and across follow-up waves. Two neurofunctional subtypes were identified within poverty-exposed youth: subtype-1 showed greater vulnerability than higher-income peers (Δβ = 0.149; P < .001), whereas subtype-2 showed attenuated vulnerability and did not differ from higher-income peers (Δβ = 0.049; P = .135); this pattern persisted longitudinally. Among poverty-exposed youth in subtype-2 with high temperamental inhibitory control, the association between cumulative ELA and later behavioral problems was no longer significant. Conclusions Childhood poverty strengthened the translation of adversity burden into later behavioral problems, but inhibitory-control profiles differentiated higher- and lower-risk pathways within poverty, highlighting inhibitory control as a candidate target for prevention.

Abstract Introduction Sleep perception impairment (SPI) in depression refers to a marked discrepancy between subjective and objective sleep, similar to paradoxical insomnia. Many depressed patients report poor sleep despite normal polysomnography (PSG), yet the neurophysiological basis of this misperception remains unclear. We examined whether EEG microarchitecture features differentiate depressed patients with SPI from those without SPI. Methods We included 63 adults with major depressive disorder and insomnia symptoms who underwent overnight PSG. SPI was defined as underestimating sleep by ≥60 minutes compared with PSG-derived total sleep time. Based on this criterion, 26 participants were classified as SPI and 37 as non-SPI. Continuous EEG measures were extracted, including total spectral power (0.5–40 Hz, μV²), an interhemispheric symmetry index (ratio of NREM spindle-band power between hemispheres), and high-frequency spectral ratios (e.g., beta/delta, gamma/delta). Group differences were examined, and multivariable logistic regression evaluated EEG predictors of SPI. Linear regression assessed associations with the magnitude of subjective–objective sleep discrepancy. Results Objective sleep duration was comparable between groups (~7 hours), yet SPI participants perceived significantly shorter sleep. SPI patients showed lower total EEG power (9730 ± 2767 μV² vs 12482 ± 4179 μV², p = 0.003) and reduced EEG symmetry (0.50 vs 0.51, p = 0.02). High-frequency activity was elevated in SPI: the N2 beta/delta ratio was higher in SPI (0.34 vs 0.26, p = 0.02), with similar increases in gamma/delta ratio (p = 0.01). In adjusted logistic models, higher total EEG power predicted lower SPI likelihood (OR ≈ 0.35 per 1000 μV², p < 0.05). A 0.01 increase in symmetry index was similarly protective (OR ≈ 0.47, p < 0.05). Higher beta/delta ratio was associated with greater SPI odds (approximately two-fold per 0.1 increment). Linear regression showed that greater EEG power related to a smaller subjective-objective sleep gap (β = –34 min, p = 0.04). Conclusion Depressed patients with SPI demonstrate distinct EEG microarchitecture marked by reduced slow-wave amplitude, elevated fast-frequency activity, and subtle interhemispheric asymmetry. This objectively “lighter” sleep pattern supports a cortical hyperarousal mechanism underlying sleep misperception. Continuous EEG features—particularly total power, symmetry, and spectral ratios—may serve as promising biomarkers to identify SPI and inform targeted interventions. Support (if any)

Despite dynamic sulcal changes during youth paralleling skill development, the link between extended postnatal development and cognition remains underexplored. This study analyzes structural MRI data from 307 children (6-14 years), with longitudinal data (inter-scan interval ~1 year) available for a subset. Results reveal widespread cortical thinning, sulcal widening, and reductions in adjusted area and depth, following a chronological gradient where earliest-forming sulci undergo the most profound change. Longitudinal remodeling, rather than baseline morphometry, predicts cognitive gains. Specifically, working memory improvements are predicted by widening of the left calcarine and posterior intralingual sulci. In contrast, attention network maturation involves global changes, though executive control is specifically linked to left intraparietal sulcus widening. Gene enrichment analysis links these changes to synaptic processes. This study advances our understanding of the association between sulcal morphometry and cognitive function, elucidating potential mechanisms underlying brain development from childhood to adolescence.

Background: Nonsuicidal self-injuries (NSSIs) are an important contributing factor to adolescent suicide, and various shared factors influence the risk of both NSSIs and suicide attempts (SAs). Both are important predictors of suicide and are part of a continuum of suicidal behaviors. Further exploration of the relationship between adolescent NSSI and SA may facilitate suicide prevention efforts. Methods: An online survey was conducted among 9,140 participants. Network analysis methods were used to explore expected influence (EI), bridge expected influence (BEI), edge weights, and differences between adolescents that have and have not attempted suicide (NSSI-SA and NSSI-NoSA, respectively). Results: Of the 9,140 participants, 7,030 completed the questionnaire, yielding a participation rate of 76.91%. Participants with at least one NSSI were retained, with 2,496 (35.50%) included in the network analysis. The strongest EI node for both networks was “emotion regulation strategies” (E = 1.389 and 1.393), and that for BEI was “personal distress” (Interpersonal Reactivity Index—personal distress; E = 0.497 and 0.492). Network comparisons revealed significant differences in NSSI 4 (“intentionally hitting walls, tables, and other hard objects”; E(Δ) = −0.384, P < 0.001), significant differences in BEI with regard to “perspective taking” (Interpersonal Reactivity Index—perspective taking; E(Δ) = −0.215, P < 0.001), and significant differences in edge weights between NSSI 4 and NSSI 5 (“intentionally hurting oneself by hitting with a fist, palm, or hard object”; E(Δr) = −0.173, P < 0.001). Conclusions: Our study suggests that interventions in the form of emotion regulation strategies can alleviate symptoms throughout the entire network. Attention should be paid to instances when NSSI 4 and NSSI 5 behaviors co-occur frequently.

BACKGROUND:The glymphatic system's role in early schizophrenia remains unclear. This study investigated glymphatic function using the DTI-ALPS index in first-episode schizophrenia (FES), exploring its relationships with clinical features, antipsychotic treatment, and cognitive function. METHODS:The study included 37 first-episode drug-naïve (FESDN) patients, 22 first-episode treated (FEST) patients, and 42 demographically matched healthy controls (HCs). ALPS indices were calculated, and clinical assessments included cognitive function (RBANS), psychiatric symptoms (PANSS), and peripheral biomarkers. RESULTS:FES patients exhibited left-hemisphere predominant reductions in ALPS indices (p < 0.01). An interaction was observed between disease duration and treatment: the FEST group had lower DTI-ALPS than the FESDN group in patients with shorter illness durations, but this group difference diminished with longer disease duration (βinter = -0.8 × 10-2, p < 0.05). The ALPS-cognition association pattern differed between patients and HCs: left ALPS was positively correlated with RBANS in HCs (βHC > 0, p < 0.01) but negatively correlated in FES patients (βFESDN = -16.6, βFEST = -21.6, p < 0.1). Peripheral white blood cell count and total bilirubin were negatively associated with left ALPS in the FEST group (WBC: β = -0.63 × 10-2, p = 0.025; TBIL: β = -0.80 × 10-2, p = 0.039), with significant group interaction effects (p < 0.05). CONCLUSION:FES patients demonstrate left-lateralized glymphatic dysfunction, the severity of which is dynamically modulated by treatment and disease duration. The observed negative ALPS‑cognition correlation raises the possibility of a compensatory mechanism in early schizophrenia. Associations noted between peripheral inflammatory/metabolic markers and ALPS offer tentative support for an interactive "periphery-brain" clearance system. DTI-ALPS may serve as a dynamic biomarker reflecting neuropathological mechanisms in early schizophrenia.

The rapid advancement of AI research automation systems--including AI Scientist, data-to-paper, and Agent Laboratory--has demonstrated the potential for autonomous scientific discovery. However, existing benchmarks for evaluating these systems focus predominantly on fundamental sciences (machine learning, physics, chemistry), overlooking the unique challenges of medical clinical research: complex survey designs, inferential statistics with confounding control, adherence to reporting standards (STROBE, CONSORT), and the requirement for clinically actionable interpretation. We present MedResearchBench, the first benchmark specifically designed to evaluate AI systems on medical clinical research tasks. MedResearchBench comprises 16 tasks spanning 7 clinical domains (cardiovascular, oncology, mental health, metabolic, respiratory, neurology, infectious disease), built on publicly available datasets (the National Health and Nutrition Examination Survey [NHANES] and the Surveillance, Epidemiology, and End Results [SEER] program) with ground truth from 16 high-quality published papers (IF range: 2.3-51.0). Each task is evaluated along 6 medical-specific dimensions: statistical methodology, results accuracy, visualization quality, clinical interpretation, confounding sensitivity, and reporting compliance. We describe the benchmark design rationale, task construction methodology, paper selection criteria with anti-paper-mill filtering, and a detailed analysis of task characteristics including methodological diversity, evaluation dimension coverage, and difficulty stratification. To demonstrate benchmark executability, we evaluate an agentic data2paper pipeline on 3 pilot tasks spanning all three difficulty tiers, achieving scores of 72/100 (Tier 1, Cardio\_000), 69/100 (Tier 2, Mental\_000), and 75/100 (Tier 3, Metabolic_002), with a mean score of 72/100 (B-level). Survey-weighted methodology was correctly implemented across all tasks; primary limitations included covariate incompleteness and reference group misspecification. MedResearchBench addresses a critical gap in AI research evaluation and provides a standardized, community-extensible platform for assessing whether AI systems can conduct clinically sound, publication-quality medical research. All task materials are publicly available at https://github.com/TerryFYL/MedResearchBench. ### Competing Interest Statement The authors have declared no competing interest. ### Funding Statement This study did not receive any funding ### 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: The study used ONLY openly available human data that were originally located at: This study used only publicly available, de-identified data from two U.S. federal sources that were openly available before the initiation of the study: National Health and Nutrition Examination Survey (NHANES), conducted by the National Center for Health Statistics (NCHS), Centers for Disease Control and Prevention. Data are freely downloadable at: https://www.cdc.gov/nchs/nhanes/ Surveillance, Epidemiology, and End Results (SEER) Program, maintained by the National Cancer Institute. Data are freely downloadable at: https://seer.cancer.gov/ No restricted-access datasets were used. All data were downloaded from the above public repositories without application, screening, or registration requirements. 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 All benchmark task materials, evaluation checklists, and variable codebooks are publicly available at https://github.com/TerryFYL/MedResearchBench (MIT License). The underlying data sources are freely downloadable from the U.S. Centers for Disease Control and Prevention (https://www.cdc.gov/nchs/nhanes/) and the National Cancer Institute (https://seer.cancer.gov/). No proprietary data were used in this study.