Background:The integration of large language models (LLMs) into medicine has reshaped health care delivery, education, and research. Although proprietary models face challenges such as data privacy, regulation, and adaptability, DeepSeek, an open-source LLM, has emerged as a customizable and cost-effective alternative with significant potential for clinical and operational applications. However, the rapid expansion of research in this area necessitates a systematic mapping of its landscape, applications, and challenges. Objective:This study combines bibliometric analysis with a scoping review to systematically map and characterize the literature on DeepSeek's medical applications. The aims were to (1) analyze publication trends, leading contributors, and research themes and (2) identify primary application domains, strengths, limitations, and future directions. Methods:Following the framework by Arksey and O'Malley and the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) guidelines, a systematic search was conducted using PubMed, Web of Science, and Scopus from January 20, 2025, to November 30, 2025. Bibliometric analysis was then used to quantify publication trends, productivity, and research themes across 371 papers. The scoping review thematically synthesized the applications, strengths, and limitations of 353 original articles. Results:The publication output showed a progressive increase, with China (n=163), Turkey (n=52), and the United States (n=48) as leading contributors. Keyword co-occurrence analysis formed 7 clusters; the 3 most frequent keywords were "large language model," "artificial intelligence," and "patient education." DeepSeek has shown promising yet preliminary performance across multiple domains, including patient education, clinical decision support, medical education, workflow optimization, and medical research. The evidence base remains predominantly low in quality, with 66.6% (235/353) of original articles classified as low-quality evidence, consisting largely of unvalidated benchmarking, simulated cases, and single-center retrospective analyses. Only 6.8% (24/353) of studies met the criteria to be considered high quality, and prospective randomized trials assessing patient-relevant outcomes were notably absent. Conclusions:Publications on DeepSeek's medical applications increased progressively from January 2025 through November 2025, with China, Turkey, and the United States as the leading contributors. The scoping review found that DeepSeek has been evaluated across 5 domains (patient education, clinical decision support, medical education, workflow optimization, and research), with variable but often competitive performance relative to proprietary models. Strengths included readability, diagnostic accuracy in select specialties, cost-efficiency, and local deployability. Limitations included inconsistent cross-specialty performance, hallucinations, ethical concerns, data privacy issues, and regulatory gaps. The evidence base is predominantly low-quality and simulation-based, with few prospective trials or randomized controlled trials. These findings indicate that DeepSeek's clinical readiness varies, and future research should address prospective validation, multimodal capabilities, bias mitigation, human oversight, and equitable access.

Precision medicine for Alzheimer’s disease (AD) requires the development of a robust management framework grounded in individualized disease staging systems. To date, only a limited number of studies have supplemented the existing AD staging systems. This retrospective study included 7491 MRI examinations from five independent cohorts. We used a novel pseudo-healthy synthesis method to capture individualized brain atrophy patterns. An individualized brain atrophy score (BAS) was computed from the 30 regions with the most severe brain atrophy and used to stratify participants into distinct disease stages. The Jenks natural breaks optimization method was used to determine an optimal number of disease stages based on the individual BAS. BAS exhibited a strong biological basis and revealed a synergistic relationship among biomarker-based staging systems. Four stages were delineated based on the BAS for participants with MCI and clinically diagnosed AD. Stage I showed a slight cognitive decline with only mild hippocampal atrophy evident. Stage II showed mild cognitive decline and mild brain atrophy and shrinkage, extending to the temporal and parietal lobes. Stage III showed moderate cognitive decline and more severe brain atrophy in the temporal lobe, amygdala, hippocampus, parietal lobe, and frontal lobe. Stage IV showed severe mental impairment and diffuse atrophy across the whole brain. The disease stages are associated with dementia severity and abnormalities in AD biomarkers, such as cerebrospinal fluid (CSF) Aβ1–42, CSF total tau, CSF p-tau181, and cognitive scores. Furthermore, those MCI participants at higher disease stages at baseline have a higher risk of progressing to clinically diagnosed AD dementia even under the A/T-negative status. The individualized staging system can accurately assess disease severity, enabling risk stratification at ultra-early pathological stages and facilitating precise AD management.

Genes impacting the bioaccumulation of perfluoroalkyl and polyfluoroalkyl substances (PFASs)and their neurotoxic effects on the brain and behavior remain unclear. Here,we examined genome-wide associations with serum accumulation of 13 PFASs in 6,823 Chinese adults. We revealed that perfluoroheptanoic acid (PFHpA) accumulation was associated with genetic variants at two loci (3q29: P = 5.20 ×10-19; 6p22.2: P = 3.69 ×10-23), mapping to 56 genes.Blood expression of 27 of these genes was associated with PFHpA accumulation in 573 subsamples. Eight genes showed potential causal effects on PFHpA accumulation,functionally linked to innate immunity (TRIM38, ZDHHC19, MUC20)and organic solute transport (SLC51A and SLC17A3). We assessed the impact of PFASs on cortical thickness and surface area, white matter fractional anisotropy and mean diffusivity,along with 25 behavioral phenotypes. We identified that seven PFASs were correlated with reduced cortical morphology, primarily in the prefrontal cortex. We also found a statistical causal effect of PFHpA accumulation on the surface area in the right frontomarginal cortex, which mediated the effect of PFHpA on anxiety. These findings indicate that serum PFHpA accumulation may be regulated by genes related to innate immunity and solute transport, heightening anxiety by impairing the prefrontal cortex.

The plateau pika (Ochotona curzoniae) profoundly alters alpine meadow ecosystems on the Qinghai-Tibet Plateau (QTP) through its bioturbation activities. A major challenge for precise monitoring and management lies in reconciling this impact. To address this, we present an advanced fine-scale analytical method that combines unmanned aerial vehicle (UAV) imagery, a deep learning (DL) model, and Generalized Additive Mixed Models (GAMMs) to unravel the distribution of pika burrows at a fine scale. The U-net semantic segmentation model accurately extracts burrows at a 30-m flight altitude (Precision = 84.00%, counting R2 = 0.86), outperforming coarser resolutions (100 m) in detecting small ecological features. GAMM analysis revealed that pika colonization is constrained by a combination of vegetation structure (NDVI) and micro-topography (aspect). Specifically, burrow density showed a negative linear response to NDVI, with a notable non-linear interaction identifying a "high-risk niche" where pikas aggregate on shady, northern slopes with sparse vegetation (NDVI <0.4) to balance thermoregulation and predator avoidance. Conversely, dense vegetation (NDVI >0.6) constitutes a strong ecological threshold, universally suppressing burrowing regardless of topography. These findings clarify the fine-scale microhabitat selection mechanisms of alpine fossorial mammals, providing a new paradigm for monitoring and managing grassland herbivores. The proposed framework supports precision ecological regulation through threshold-based approaches, positioning vegetation restoration as a central management strategy.

Excessive exercise can induce metabolic disturbances that precede overt clinical disease. As the central metabolic organ, the liver plays a pivotal role in systemic metabolic adaptations to exercise, although its dynamic metabolic response remains poorly characterized. Characterizing hepatic metabolic shifts could advance early diagnostic and preventive strategies. Single exhaustive exercise (SEE) is an acute, short-duration, high exercise load, often used to model a single bout of supra-physiological exertion, whereas repeated exhaustive exercise (REE) models the cumulative physiological stress induced by consecutive exhaustive exercise sessions. This study systematically delineates temporal metabolic alterations in murine liver following SEE and REE. C57BL/6J mice were subjected to a single bout of exhaustive exercise or daily REE regimens for 7 consecutive days. Liver tissues and serum samples were collected at predetermined intervals (0, 1, 6, 12, 24, 48 h post-exercise) for comprehensive analysis, including untargeted metabolomics, histopathological evaluation, and quantification of liver injury biomarkers. SEE provoked transient metabolic perturbations that resolved within 24 h, whereas REE induced progressive metabolic remodeling, particularly involving amino acid metabolism. Both exercise modalities caused histologically confirmed hepatic injury, and the biomarkers for liver injury were elevated at an early stage but recovered within 24 h. Multivariate analysis identified "steroid hormone biosynthesis" and "taurine/hypotaurine metabolism" as key modules correlating with injury severity. Time-series analysis showed that most injury-related metabolites in the SEE group returned to baseline, whereas those in the REE group remained elevated through 48 h, suggesting sustained metabolic alterations within the observation window, which may reflect delayed recovery and/or adaptive metabolic remodeling in response to repeated exhaustive exercise. Our findings reveal distinct patterns of hepatic metabolic alteration: acute exhaustive exercise triggers self-limited metabolic adjustments, whereas repeated exhaustive exercise induces more sustained metabolic remodeling. These results underscore the importance of personalized exercise regimens and suggest that modulation of specific metabolic pathways may represent a potential strategy for mitigating exercise-induced hepatic stress.

INTRODUCTION: Mild cognitive impairment (MCI), a prodromal stage of Alzheimer's disease (AD), shows pronounced clinical heterogeneity poorly explained by pathology burden, representing a gap complicating prognosis. As the brain operates as a complex network for information integration, we hypothesized that connectome architecture mediates the link between AD pathology and clinical expression. METHODS: We developed a framework integrating structural and functional connectomes from multi-center cohorts, performing connectome-based subtyping in MCI, with analyses of upstream pathology, downstream phenotypes, and transcriptomic associations. RESULTS: This approach identified an "MCI-compromised" (MCI-C) subgroup characterized by extensive structural-functional connectomic disruption and an "MCI-preserved" (MCI-P) subgroup with relatively preserved connectome integrity. Despite comparable pathology, MCI-C demonstrated more severe neurodegeneration, accelerated cognitive decline, and elevated progression risk. Multiscale analyses linked these patterns to transcriptomic profiles of mitochondrial, synaptic, and neuroimmune processes. DISCUSSION: These findings demonstrate that the connectome acts as a critical mediator, rather than a passive endophenotype, shaping AD clinical expression.

Neuroanatomical heterogeneity in Alzheimer's disease (AD) hinders precision diagnosis and treatment, as distinct brain phenotypes may correspond to different disease subtypes. However, MRI-based subtype classifications are often confounded by co-occurring pathologies and non-AD factors, such as genetic predisposition and environmental influences, limiting their clinical interpretability. We propose 3D-DisAD, an unsupervised deep learning framework that disentangles AD-specific neuroanatomical variations from unrelated influences and clusters patients into subtypes with homogeneous brain phenotypes. The framework comprises two synergistic networks: (1) Contrastive Disentanglement Network, which separates AD-specific variations from those shared by AD patients and healthy controls; and (2) Transformation Generation Network, which refines these disease-specific variations by transforming healthy brain representations into realistic, pathology-consistent anatomies via diffusion-based generative modeling. Evaluated on four public datasets, 3D-DisAD reveals strong correlations between the disentangled AD-specific variations and diverse clinical and biological profiles, validating their relevance. Using these variations, we identify four AD subtypes with significant differences in biomarkers, cognitive trajectories, and genetic signatures, and uncover distinct longitudinal progression patterns that suggest potential windows for early intervention. By disentangling AD-specific variations, our method enables more precise patient stratification and personalized treatments, particularly in the early stage of AD. Code is available at: https://github.com/cnuzh/3D-DisAD.

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.

Background Alzheimer’s disease (AD) is increasingly conceptualized as a disconnection syndrome involving widespread alterations in large-scale brain networks. Previous studies using morphometric similarity networks (MSNs) have revealed broad structural and transcriptomic changes, yet vertex-level structural disconnection and its molecular basis remain poorly understood. We applied morphometric inverse divergence (MIND), an innovative approach for fine-grained mapping of structural disconnection and its transcriptomic correlates in AD. Methods Utilizing two independent datasets: [ADNI (219 AD, 219 cognitively normal, CN) and the Qilu dataset (100 AD, 137 CN)], we mapped robust MIND network alterations in AD patients and examined their associations with cognitive performance and biomarker quantifications. Additionally, we linked MIND connectome to spatial gene expression using partial least squares regression, followed by gene enrichment analysis to identify relevant biological pathways. Finally, to validate the clinical utility of MIND, a residual deep neural network (ResDNN) was developed to compare its diagnostic performance against MSNs in distinguishing AD from CN. Results Significantly decreased MIND degree was identified in the bilateral frontal, lateral occipital, and posterior temporal lobes (P FDR < 0.05), positively correlating with MMSE score and FDG-PET SUVR (all P < 0.001). Conversely, increased MIND degree was observed in the bilateral cuneus, entorhinal, lingual, and parahippocampal regions (P FDR < 0.05), negatively correlating with cognition assessment, CSF Aβ-42 levels and FDG-PET SUVR (all P < 0.001). These AD-related MIND alterations were spatially correlated with gene expression profiles crucial for synaptic function, neurotransmission, and metabolic regulation. Importantly, MIND achieved superior diagnostic efficacy (AUC=0.90/0.88 in ADNI/Qilu) over MSNs. Conclusions We mapped a robust pattern of structural disconnection in Alzheimer's disease with MIND approach and associate it with particular transcriptomic signatures. These findings not only improve our mechanistic understanding of AD as a disconnection syndrome but also demonstrate MIND as a sensitive tool for identifying disease-specific alterations, holding promise for future mechanistic and clinical investigations into AD pathology.

Photoperiod changes serve as critical environmental signals for seasonally reproducing animals, with their decoding dependent on the precise functioning of endogenous circadian clocks. Here we conducted a comprehensive analysis of diurnal transcripts and metabolites profiles in the testes of Brandt's voles (Lasiopodomys brandtii) under long photoperiod (LP) and short photoperiod (SP), to elucidate the role of circadian rhythms within peripheral testicular tissue in regulating reproduction. Phenotypic results demonstrated that LP promoted testicular development, whereas SP suppressed it. Then we observed relatively weaker transcriptional and metabolic rhythmicity in the testes under both photoperiods. The function of diurnal rhythmic genes was mainly related to homeostasis, developmental, reproductive behavior, motility and sperm flagellum assembly pathways, suggesting that diurnal clocks act as photoperiodic timekeeper for reproductive timing. In contrast, testosterone biosynthesis and key energy metabolism pathways essential for spermatogenesis were largely governed by non-rhythmic genes expression patterns. Together, these findings underscore the non-uniform pervasiveness of diurnal rhythmicity in testis, where reproductive initiation driven by the diurnal clock, while subsequent spermatogenic processes progressively deviate from 24 h rhythmicity.

BACKGROUND:As an extension of diffusion tensor imaging (DTI), diffusion kurtosis imaging (DKI) quantifies non-Gaussian water diffusion and has been applied to explore brain disorders. However, the genetic architecture of brain DKI phenotypes remains unknown. METHODS:Here, we estimated heritability and conducted genome-wide association studies (GWASs) for 804 DKI phenotypes across 188 brain structures in 4183 participants. To determine whether DKI-GWASs provides genetic insights beyond DTI-GWASs, we compared results from 804 DKI-GWASs and 752 DTI-GWASs in the same cohort. To clarify the biological significance of DKI phenotypes, we examined associations between DKI phenotypes and brain health-related outcomes within the CHIMGEN, and explored associations between polygenic risk scores (PRSs) of DKI phenotypes and mental disorders in the UK Biobank. FINDINGS:Of 804 DKI phenotypes, 275 showed significant heritability (P < 0.05; h2 range: 0.143-0.602). We detected 280 significant associations (P < 5 × 10-8), with 38 surviving Bonferroni correction (P < 1.54 × 10-10). These associations were unevenly distributed across chromosomes, DKI phenotype subgroups, and brain structures. Among 229 independent variant-structure associations for DKI, 175 (76.4%) were DKI-specific. We observed 930 associations between DKI phenotypes and brain health-related outcomes (P < 0.05; ten Bonferroni-significant with P < 1.02 × 10-5), and 200 between PRSs and mental disorders (P < 0.05; one Bonferroni-significant with P < 9.61 × 10-5). INTERPRETATION:This study delineates the genetic architecture of brain DKI phenotypes, identifies complementary genetic insights into brain microstructure, and provides biologically relevant endophenotypes for investigating neural mechanisms underlying brain disorders. FUNDING:National Natural Science Foundation of China, National Key Research and Development Program of China, Tianjin Key Medical Discipline Construction Project, and Tianjin Natural Science Foundation.

Cerebral asymmetry is a core principle of human brain organization, showing dynamic changes across the lifespan and alterations in brain disorders. However, it remains unclear whether lifespan trajectories of asymmetry differ across populations. We compared lifespan structural asymmetry normative charts of 221 cerebral imaging phenotypes from 43,037 Chinese and 56,339 Western participants aged 0–100 years. The two populations showed distinct lifespan asymmetry patterns in 26.2% of the phenotypes. Chinese-minus-Western asymmetry difference curves displayed distinct patterns across brain phenotypes: rightward (45.7%), leftward (26.2%), rightward-to-leftward (11.8%), leftward-to-rightward (10.0%), and unclassified (6.3%). Population-matched normative models outperformed population-unmatched normative models in capturing normal asymmetry variability among healthy individuals and in detecting abnormal asymmetry deviations in patients with Alzheimer’s disease, mild cognitive impairment, schizophrenia, and major depressive disorder. These findings indicate that population mismatch can bias chart-based individual-level asymmetry assessment and underscore the need for population-representative brain asymmetry normative charts.

Purpose To examine common patterns among different computer-aided diagnosis (CAD) models for Alzheimer disease (AD) using structural MRI data and to characterize the clinical and imaging features associated with their misclassifications. Materials and Methods This retrospective study used 3258 baseline structural MRI scans from five multisite datasets and two multidisease datasets collected between September 2005 and December 2019. The 3D Nested Hierarchical Transformer (3DNesT) model and other CAD techniques were used for AD classification using 10-fold cross-validation and cross-dataset validation. Subgroup analysis of CAD-misclassified individuals compared clinical and neuroimaging biomarkers using independent t tests with Bonferroni correction. Results This study included 1391 patients with AD (mean age, 72.1 years ± 9.2 [SD]; 757 female), 205 with other neurodegenerative diseases (mean age, 64.9 years ± 9.9; 117 male), and 1662 healthy controls (mean age, 70.6 years ± 7.6; 935 female). The 3DNesT model achieved 90.0% ± 2.3 cross-validation accuracy and 82.2%, 90.1%, and 91.6% accuracy in three external datasets. Further analysis suggested that the false-negative subgroup (n = 223) exhibited minimal atrophy and better cognitive performance on the Mini-Mental State Examination (MMSE) than the true-positive subgroup (MMSE score in false-negative subgroup, 21.4 ± 4.4; true-positive subgroup, 19.7 ± 5.7; P value family-wise error [PFWE] < .001), despite displaying similar levels of amyloid β (false-negative subgroup, 705.9 pg/mL; true-positive subgroup, 665.7 pg/mL; PFWE = .99) and tau (false-negative subgroup, 352.4 pg/mL; true-positive subgroup, 371.0 pg/mL; PFWE = .99) burden. Conclusion A subgroup of patients with false-negative classification for Alzheimer disease exhibited atypical structural MRI patterns and clinical measures, fundamentally limiting the diagnostic performance of CAD models based solely on structural MRI. Keywords: MR Imaging, Dementia, Computer Applications-3D, Alzheimer's Disease, Computer-aided Diagnosis, Misclassification, Atypical AD Supplemental material is available for this article. © RSNA, 2025 See also commentary by Nasrallah in this issue.

The mechanism by which the increasing environmental challenge of urbanicity impacts the brain, personality and mental disorders remains unclear. Here, grounded in life history theory, we tested the hypothesis that age at menarche (AAM) mediates the effects of early-life urbanicity on adult regional brain volumes and personality traits associated with mental disorders. In a sample of 2,950 young Chinese women, we discovered that higher levels of early-life urbanicity were associated with earlier AAM, which in turn correlated with reduced medial prefrontal volume and lower levels of agreeableness and reward dependence in adulthood. Urbanicity-related factors, particularly family socioeconomic status, also influenced these neurobehavioral traits through AAM. The urbanicity- and AAM-related brain and personality traits were changed in patients with major depressive disorder and schizophrenia. These findings suggest that life history theory may serve as a mechanism through which early-life urbanicity influences the adult brain and personality traits associated with mental disorders in women. It is unclear how early-life urbanicity influences adult neurobehavioral traits. This study reveals that earlier menarche mediates the relationship between early-life urbanicity and adult neurobehavioral traits associated with mental disorders.

OBJECTIVE:Previous studies examining post-stroke aphasia (PSA) patients via resting-state functional magnetic resonance imaging (rs-fMRI) have predominantly focused on static functional connectivity. In contrast, the current investigation aims to elucidate the alterations in dynamic functional network connectivity (dFNC) among PSA patients. METHODS:We recruited 40 PSA patients and 41 age-, gender-, and education-matched normal controls (NCs). The participants underwent rs-fMRI and the Western Aphasia Battery (WAB) test. Independent component analysis (ICA) was used to identify resting-state networks (RSNs), and dFNC was constructed using a sliding window technique followed by k-means clustering to classify distinct dynamic network states. We compared the dFNC differences between the PSA and NC groups and examined their relationships with clinical outcomes. RESULTS:The dynamic analysis identified 4 distinct dFNC states: state 1 (modular network state), state 2 (pan-network hyperconnectivity state), state 3 (intra-network coordinated state), and state 4 (sparse connectivity state). Notable group differences were observed: compared with NCs, the PSA group demonstrated significantly reduced connectivity within the language network (LN) and increased connectivity between the cerebellar network (CN) and the default mode network (DMN) in state 3; significantly reduced connectivity changes within the LN and between the LN and executive control network (ECN) / CN were noted in state 4. Additionally, fraction time and mean dwell time in the modular network state were positively correlated with the WAB score. INTERPRETATION:Alteration in dFNC could serve as a sensitive biomarker toward language impairment in PSA patients, with implications for diagnostic assessment and therapeutic intervention planning.

Alzheimer’s disease (AD) is marked by disrupted brain network connectivity, which impairs functional hierarchy and contributes to cognitive decline. Two fundamental questions, however, remain open: what specific changes occur in the hierarchical organization of the AD brain, and whether rectifying these changes can restore cognitive function. To answer these, we first analyzed individualized functional hierarchical architecture across three large, independent fMRI datasets (MCADI, N = 711; ADNI, N = 621; OASIS-3, N = 506). We identified a reproducible pattern of hierarchical remodeling in AD, characterized by expansion of the dorsal attention network A and shrinkage of the control network A, with spatial variability shaped by underlying brain tissue properties. To evaluate the therapeutic relevance of this signature, we conducted a randomized controlled trial of transcranial alternating current stimulation (tACS; N=44). Targeted stimulation selectively reversed the remodeling trajectory, suppressing dorsal attention network expansion and countering control network contraction. These network improvements persisted for three months and were accompanied by sustained cognitive gains, with 80% of participants showing measurable improvement. Our results reveal a functional hierarchical signature of AD and establish its potential as a novel interventional target, while also providing mechanistic insights into the action of non-invasive neuromodulation.

With the worldwide increase in only-child families, it is crucial to understand the effects of growing up without siblings (GWS) on the adult brain, behaviour and the underlying pathways. Using the CHIMGEN cohort, we investigated the associations of GWS with adult brain structure, function, connectivity, cognition, personality and mental health, as well as the pathway from GWS to GWS-related growth environments to brain and to behaviour development, in 2,397 pairs of individuals with and without siblings well matched in covariates. We found associations linking GWS to higher language fibre integrity, lower motor fibre integrity, larger cerebellar volume, smaller cerebral volume and lower frontotemporal spontaneous brain activity. Contrary to the stereotypical impression of associations between GWS and problem behaviours, we found positive correlations of GWS with neurocognition and mental health. Despite direct effects, GWS affects most brain and behavioural outcomes through modifiable environments, such as socioeconomic status, maternal care and family support, suggesting targets for interventions to enhance children's healthy growth.