Accurate sleep stage classification based on single-channel electroencephalogram (EEG) signals remains a fundamental challenge in automated sleep staging. Existing deep learning architectures may be limited by fixed receptive fields, sequential processing bottlenecks, and inadequate adaptation to inter-subject variability. We propose an Adaptive Multi Scale Temporal-Spectral Network (AMT-SNet) designed for efficient and accurate single-channel EEG-based sleep stage classification. Our method employs a dual-branch architecture with temporal-spectral feature extraction, subject-adaptive calibration, and component-wise training to achieve robust and personalized sleep stage classification. We evaluated our method on the Sleep-EDF-20 dataset for five-class sleep stage classification and further validated its generalization ability using the Sleep-EDF-78 dataset. The results on the Sleep-EDF-20 dataset demonstrate that AMT-SNet outperforms eight state-of-the-art models, achieving an accuracy of 85.8

Group independent component analysis (ICA) has been extensively used to extract brain functional networks (FNs) and associated neuroimaging measures from multi-subject functional magnetic resonance imaging (fMRI) data. However, the inherent noise in fMRI data can adversely affect the performance of ICA, often leading to noisy FNs and hindering the identification of network-level biomarkers. To address this challenge, we propose a novel method called group information-guided smooth independent component analysis (GIG-sICA). Our method effectively generates smoother functional networks with reduced noise and enhanced functional coherence, while preserving intra-subject independence and inter-subject correspondence of FN. Importantly, GIG-sICA is capable of handling different types of noise either separately or in combination. To validate the efficacy of our approach, we conducted comprehensive experiments, comparing GIG-sICA with traditional group ICA methods on both simulated and real fMRI datasets. Experiments on five simulated datasets, generated by adding various types of noise, demonstrate that GIG-sICA produces smoother functional networks with enhanced spatial accuracy. Additionally, experiments on real fMRI data from 137 schizophrenia patients and 144 healthy controls demonstrate that GIG-sICA more effectively captures functionally meaningful brain networks and reveals clearer group differences. Overall, GIG-sICA produces smooth and precise network estimations, supporting the discovery of robust biomarkers at the network level for neuroscience research.

BACKGROUND:The difference between neuroimaging-predicted brain age and chronological age, the predicted age difference (PAD), has been studied as a potential biomarker reflecting individual brain health. Although previous large-scale studies have shown that brain age deviations occur across multiple disorders, cross-disorder comparisons of PAD within a unified framework, together with identification of the neuroimaging features associated with these differences and their related gene expression profiles, remain limited. Our aims are to systematically compare brain aging across multiple common brain disorders and explore the brain patterns and biological processes underlying these differences. METHODS AND FINDINGS:In this study, structural MRI data from 45,900 healthy controls (HCs) and 2,698 patients with developmental disorders (attention-deficit/hyperactivity disorder [ADHD] and autism spectrum disorder [ASD]), addiction (alcohol use disorder [AUD], tobacco use disorder [TUD], and AUD&TUD-A&TUD), dementia (Alzheimer's disease [AD], and mild cognitive impairment [MCI]) or other psychiatric disorders (schizophrenia [SZ], bipolar disorder [BP], and major depressive disorder [MDD]), were collected to generate PAD, along with transcriptome data. Then, we calculated the PAD difference between patient and HC as Cohen's d effect sizes, derived from a linear model that accounted for age, age2, sex, and site, and further identified the interpretable brain patterns associated with the PAD difference for each diagnostic group. Finally, enrichment analyses was conducted to identify the biological function of genes relatively over- or underexpressed in association with these patterns. Results showed that while PAD was consistently greater across disorders, different brain disorders showed different degrees of abnormality, the highest effects in dementia (AD: d = 0.97, 95% confidence interval (CI) [0.82,1.13]; p < 0.001 and MCI: d = 0.45, 95% CI [0.34,0.56]; p < 0.001), followed by addiction (A&TUD: d = 0.84, 95% CI [0.44,1.23]; p < 0.001, TUD: d = 0.72, 95% CI [0.49,0.96]; p < 0.001, and AUD d = 0.62, 95% CI [0.39,0.84]; p < 0.001) and psychiatric disorders (SZ: d = 0.53, 95% CI [0.30,0.76]; p < 0.001, BP: d = 0.46, 95% CI [0.22,0.69]; p < 0.001 and MDD: d = 0.28, 95% CI [0.11,0.46]; p < 0.001), but not different from expected in developmental disorders (ASD: d = 0.06, 95% CI [-0.04,0.16]; p = 0.36) and ADHD: d = 0.01, 95% CI [-0.14,0.15]; p = 0.98). Furthermore, higher PAD values in patient groups were linked to specific spatial brain patterns, including the frontotemporal network in psychiatric disorders, default mode network-salience network-putamen-thalamus in addiction and fronto-occipital network in dementia. Prefrontal cortex involvement was common across disorders, and disorder-specific brain patterns associated genes were enriched in different biological processes. A limitation of our study is that psychiatric disorders and addiction have high comorbidity, and these potential confounders were not considered. CONCLUSIONS:In summary, the different brain aging patterns, each based around specific underlying circuits, may serve as neuroimaging biomarkers for understanding the neural aging mechanisms in commonly occurring brain disorders. Future studies should test whether these disorder-specific brain aging patterns can serve as useful biomarkers to guide critical clinical decision-making.

Dynamic functional connectivity (dFC) analysis investigates how the functional interactions between brain regions change over time by identifying recurring connectivity patterns, known as dFC states, and tracking transitions between them. Non-negative matrix factorization (NMF) has been used in dFC analysis because it produces non-negative dFC states and coefficients, interpreting dFC states and their transitions straightforwardly. However, existing NMF-based methods are limited to processing dFC data with exclusively positive values, failing to align with the functional correlations and anti-correlations between brain regions. This paper proposes an orthogonal semi-nonnegative matrix factorization (OSemiNMF) method, extending NMF to directly handle mixed-sign dFC data. Furthermore, an orthogonality constraint on the bases (i.e., dFC states) is incorporated to enhance the uniqueness of dFC states. For 10 simulated datasets with varying properties, our method outperforms comparison methods, supporting its superior ability to capture dFC states and state transitions. Using four resting-state fMRI datasets consisting of 708 healthy controls (HCs) and 537 schizophrenia patients (SZs), our method identifies reproducible dFC states and state transitions across datasets. Further, our findings reveal that SZs spend less time in high-connectivity states compared to HCs. Our study identifies meaningful and reproducible biomarkers of schizophrenia, mainly involving the connectivity associated with the sub-cortical domain. In summary, the OSemiNMF method facilitates the dFC analysis for understanding brain dynamics.

Background Anxiety and depression share many symptoms, frequently leading to diagnostic uncertainty. They are also the most common co-occurring psychiatric disorders. Standard clinical evaluations typically diagnose each condition categorically rather than assessing their relative severity when both are present. Prior imaging studies largely concentrated on delineating similarities and distinctions between the two.Methods To address this gap, we propose a novel noisy label learning method to investigate the biotypes and relevant neural characteristics among anxiety, depression, and their comorbidity. Our approach constructs a robust classification model to address potential diagnostic confusion among anxiety, depression, and comorbidity using a noisy label learning strategy, and employs the trained model to identify the anxiety and depression biotypes, as well as comorbidity biotypes characterized by differing degrees of anxiety and depression predominance.Results Using brain functional network connectivity (FNC) from 502 depression patients, 245 anxiety patients, 177 comorbid patients with both anxiety and depression, and 500 healthy controls, we not only identify four distinct biotypes, but also reveal neural linkages between comorbid biotypes and depression and anxiety biotypes. Compared to the depression biotype (Biotype 1), the anxiety biotype (Biotype 2), and the comorbid biotype with an anxiety predominance (Biotype 3-2) demonstrate comparable alterations across seven FNCs, mainly including the connections between the cognitive control and visual domains. Relative to the anxiety biotype (Biotype 2), the depression biotype (Biotype 1), and the comorbid biotype with a depression predominance (Biotype 3-1) exhibit similar connectivity profiles across 16 FNCs, primarily involving connections within the cognitive control domain, as well as between the sensorimotor and visual domains. Furthermore, biotypes are characterized by distinct patterns of behavioral symptoms that are consistent with their underlying neural relationships.Conclusions In summary, we propose a neuroimage-based model that addresses the diagnostic ambiguity between anxiety and depression and derive data-driven subtypes for promoting the precise diagnosis.

We propose TB-GCAN, a tri-branch cross-attention graph neural network for schizophrenia classification using multimodal MRI, including sMRI, fMRI, and DTI. Built on a multi-site dataset of 1191 samples from seven scanning sites, the model exploits atlas-defined one-to-one anatomical correspondence across modalities to enable node-level cross-attention during intermediate representation learning. In 7-site leave-one-site-out evaluation, TB-GCAN achieved 84.63% accuracy and outperformed GAT, GCN, CNN, SVM, and MMGNN in the tri-modal setting. Attention-based region ranking highlighted biologically plausible schizophrenia-related regions, and downstream analyses linked the learned imaging representations to PANSS dimensions and transcriptional programs. Unlike generic multimodal GNNs that learn cross-modal relations from data, TB-GCAN directly leverages atlas-aligned regional correspondence to perform anatomically constrained node-level interaction. These findings indicate that anatomically grounded node-level multimodal fusion can improve classification performance while preserving neurobiological interpretability, thereby providing a principled framework for multimodal schizophrenia classification and biomarker discovery.

Recent advancements in multiple kernel learning-based graph clustering methods have demonstrated significant promise in effectively learning a consensus kernel matrix from various candidate kernel matrices. This approach enables the creation of a low-dimensional representation in the nel space of high-dimensional data through self-expressiveness. However, a key challenge remains in capturing the latent geometric properties embedded within different kernel matrices hance data representation. In this paper, we propose a novel method called joint consensus learning and adaptive hypergraph regularization for graph-based clustering (JKHR). Our approach integrates an innovative adaptive hypergraph Laplacian regularizer, which is characterized fusion of multiple nearest neighbor kernel graphs, into the multiple kernel learning-based clustering framework. JKHR jointly and adaptively optimizes both the consensus kernel and the hypergraph Laplacian regularizer to achieve a low-dimensional representation that tively preserves the intrinsic geometry of the data. Experimental results on both synthetic and benchmark datasets demonstrate that JKHR outperforms state-of-the-art self-expressiveness graph clustering methods as well as traditional clustering techniques.

The subjective nature of diagnosing mental disorders complicates achieving accurate diagnoses. The complex relationship among disorders further exacerbates this issue, particularly in clinical practice where conditions like bipolar disorder (BP) and schizophrenia (SZ) can present similar clinical symptoms and cognitive impairments. To address these challenges, this paper proposes a mutualistic multi-network noisy label learning (MMNNLL) method, which aims to enhance diagnostic accuracy by leveraging neuroimaging data under the presence of potential clinical diagnosis bias or errors. MMNNLL effectively utilizes multiple deep neural networks (DNNs) for learning from data with noisy labels by maximizing the consistency among DNNs in identifying and utilizing samples with clean and noisy labels. Experimental results on public CIFAR-10 and PathMNIST datasets demonstrate the effectiveness of our method in classifying independent test data across various types and levels of label noise. Additionally, our MMNNLL method significantly outperforms state-of-the-art noisy label learning methods. When applied to brain functional connectivity data from BP and SZ patients, our method identifies two biotypes that show more pronounced group differences, and improved classification accuracy compared to the original clinical categories, using both traditional machine learning and advanced deep learning techniques. In summary, our method effectively addresses the possible inaccuracy in nosology of mental disorders and achieves transdiagnostic classification through robust noisy label learning via multi-network collaboration and competition.

Aging has important impacts on both the function and structure of the brain, yet the interplay between these changes remains unclear. Here, we present a unified framework including both single-modal and multimodal age predictions using a large UK Biobank dataset (27,793 healthy subjects, 49 to 76 years) to identify and validate brain functional network connectivity (FNC) and gray matter volume (GMV) changes associated with aging, then propose a novel analysis method to reveal various joint aging patterns, and finally investigate the association between joint function-structure changes and cognitive declines. Multimodality outperforms single modality in the age prediction, underscoring the significance of multimodal aging-related changes. Aging primarily induces synergistic changes, with both FNC and GMV decreased in the cerebellum, frontal pole, paracingulate gyrus, and precuneus cortex, indicating consistent degeneration in motor control, sensory processing, and emotional regulation, and contradictory changes with increased FNC magnitude but decreased GMV in the occipital pole, lateral occipital cortex, and frontal pole, acting as a compensatory mechanism as one ages to preserve visual acuity, cognitive ability, and behavioral modulation. Particularly, joint changes, with both FNC and GMV decreased in the crus I cerebellum and the paracingulate gyrus, show a strong Pearson correlation with the reaction time. In summary, our study unveils diverse joint function-structure changes, providing strong evidence for understanding distinct cognitive deteriorations during aging.

Brain functional network (FN) extraction is fundamental to advancing our understanding of brain function, providing critical insights into the neural mechanisms underlying cognition and behavior. Data-driven FN analysis methods have been developed to analyze functional magnetic resonance imaging (fMRI) data. However, to ensure cross-subject correspondence, group-level analyses of these methods sacrifice subject-specific variation. This trade-off between group-level alignment and subject-specific discrepancies hinders the accurate characterization of individual brain FNs. In this study, we propose a multi-subject orthogonal sparse matrix decomposition method without the need for group-level analysis, which simultaneously extracts both group-level FNs and individual FNs with cross-subject correspondence. We introduce a novel quasi-orthogonality constraint that enhances the linear independence of FNs, ensuring effective extraction of FNs, while enabling precise control over FN spatial scale. Additionally, by further incorporating a sparsity constraint, our method effectively minimizes spatial overlap between FNs, resulting in sparse representations. For simulated datasets, our method outperforms comparison methods, supporting its low parameter sensitivity and superior ability to extract FNs and time courses. Application to multi-site fMRI datasets, comprising 233 healthy controls (HCs) and 205 schizophrenia patients (SZs), validates the reproducibility of FNs extracted by our method. The results underscore the method's ability to preserve both cross-subject correspondence and individual variability. Overall, our method advances fMRI analytic capabilities by reconciling population-level consistency with individualized neural signatures, offering enhanced discriminative power for investigating neuropsychiatric disorder mechanisms and brain function.

Although foundation models have advanced many medical imaging fields, their absence in neuroimage analysis limits progress in neuroscience and clinical practice. Brain functional connectivity (FC) analysis is central to understanding brain function and widely used in neuroscience. We propose a foundation model tailored for brain functional connectivity networks (FCN). Our graph transformer model integrates node and edge embeddings to extract FCN features and adapts flexibly to classification, regression, and clustering via task-specific adapters. We validate the model on fMRI data from 10,718 scans across multiple tasks: gender classification, mental disorder classification (distinguishing schizophrenia or autism from healthy population), brain age prediction, and depressive and anxiety disorder biotyping. Compared to 14 competing methods, our model consistently outperforms them. Moreover, it facilitates biomarker discovery by identifying task-specific FC patterns. In summary, we present a novel, versatile foundation model for FCN that advances neuroimaging research through scalable and interpretable analysis.

Electroencephalography (EEG) is a crucial physiological signal that reflects real-time brain activity and exhibits inherent individual variability, making it a promising modality for identity recognition. However, the nonstationarity and complexity of EEG signals make it difficult for traditional feature extraction methods to achieve efficient signal characterization in identity recognition. To address these issues, we propose a novel spatial fractional-domain (SFD) feature extraction algorithm that retains critical spatial information through common spatial pattern (CSP) while leveraging the fractional Fourier transform (FRFT) to capture both temporal and frequency characteristics. The fractional-domain allows for flexible representation of signal features by adjusting the transformation order, thereby improving the algorithm's adaptability to the nonstationary nature of EEG signals. In addition, we have curated a new short-term speech-induced EEG dataset focusing on four primary emotions (happiness, sadness, anger, and surprise), alongside simultaneous speech signal recordings to monitor the concentration status of the subjects. Experimental results demonstrate that the proposed method achieves optimal identity recognition performance, with an accuracy peak at a fractional order of 0.1 for this dataset. Furthermore, validations on widely recognized public datasets SEED, DEAP, and FACED, showed a classification accuracy peak at fractional orders of 0.2, 0.4, and 0.1, respectively, further validating the generalizability and robustness of the SFD algorithm. These findings underscore the algorithm's effectiveness in addressing the complexities of EEG-based identity recognition.

Despite considerable efforts to uncover the neural basis of psychiatric disorders using neuroimaging, few methods utilize intrinsic brain-derived knowledge, leading to limited specificity and discriminability in biomarker identification. To leverage the inherent characteristics within the brain, we propose a prior-knowledge-guided feature selection method to flexibly unveil discriminative and target-oriented biomarkers of psychiatric disorders. Specifically, we construct a constrained sparse regularization allowing for the flexible integration of diverse prior knowledge to identify sparse neuroimaging features linked to specific psychopathology. Additionally, we simultaneously integrate graph-based regularization and redundancy-removal regularization to further ensure the discriminability and independence among the selected features. Different priors hold varying significance in identifying specific biomarkers. Four functional magnetic resonance imaging (fMRI) datasets from 708 healthy controls and 537 schizophrenia patients are used to evaluate our method integrated with various prior knowledge, revealing specific schizophrenia-related brain abnormalities. Compared with nine advanced feature selection methods, our method improves mean classification accuracy by 3.89% to 11.24%, particularly revealing reduced interactions within the visual domain and between subcortical and visual domains in schizophrenia patients. The proposed method offers flexible and precise biomarker identification tailored to specific targets, advancing the understanding and diagnosis of psychiatric conditions.

Brain age prediction using neuroimaging data is crucial for understanding the mechanisms of brain aging. However, many previous methods for predicting brain age have relied on chronological age as the guiding label for model training. The models produced may not accurately predict true brain age. To overcome this problem, we propose a new noisy label learning method to train a robust brain age prediction model by addressing the possible label bias in age. The method incorporates two deep neural networks and their corresponding mean-teacher networks, which effectively identify and leverage samples with clean and noisy labels that show consistency and inconsistency between brain age and chronological age, respectively. Experiment using simulated datasets including 10,000 samples demonstrates high predictive accuracy, with a correlation coefficient greater than 0.9 between the predicted brain age and the simulated brain age. Compared with traditional regression algorithms, our method obtains better performance for the simulated datasets. We applied our method to brain functional connectivity data of 10,000 healthy controls from the UK Biobank dataset to validate the capability of our method. Results show that the predicted brain ages have greater correlations with cognitive measures compared to the chronological age. In summary, our new noisy label learning method can effectively predict brain age using neuroimaging data by overcoming potential age bias.

Dynamic functional connectivity (dFC) analysis has revealed that functional connectivity fluctuates over short timescales, reflecting the intrinsic transitions of brain among multiple states. However, dFC data typically exhibit the characteristics of high dimensionality and noise, making it difficult to extract stable and accurate states. Furthermore, accurately identifying model order (i.e., number of states) is challenging due to lack of prior knowledge. To address the above issues, we propose a model order-free method for extracting stable states. Our method can simultaneously capture multi-scale state information and improve the stability of the state. Furthermore, our method estimates the number of states adaptively based on data-driven methods. Based on synthetic data, we evaluated the effectiveness of our method. The results showed that, compared to traditional methods, our method not only accurately estimated the number of states but also extracted states with greater robustness and precision. Additionally, we evaluated the effectiveness and stability of the method using fMRI data from 602 healthy controls and 519 schizophrenia patients. Results demonstrated that our method exhibited significant consistency among the states extracted by multiple runs. Moreover, we identified reliable biomarkers for schizophrenia. In conclusion, we propose a novel state extraction method that does not rely on predefined state numbers, while accurately and stably identifying states.

Background: Conflicting findings exist regarding the association between maternal serum zinc and neonatal birth weight. This study aimed to explore the association between maternal serum zinc and birth weight, and whether this association was modified by neonatal SOD2 polymorphism and promoter methylation. Methods: We recruited 464 mother-newborn pairs at Houzhai Center Hospital from January 2010 to January 2012. Maternal serum zinc concentration was determined using atomic absorption spectrophotometry. Neonatal SOD2 polymorphism and promoter methylation were measured by TaqMan probe assay and real-time quantitative methylation-specific PCR (QMSP), respectively. Relationships among maternal serum zinc, neonatal SOD2 promoter methylation, and birth weight were analyzed by generalized linear model (GLM). Stratified and interaction analyses were conducted to explore the modification of neonatal SOD2 polymorphism and promoter methylation on the association between maternal serum zinc and birth weight. Results: Our findings revealed that higher maternal zinc concentrations were associated with decreased birth weight (P-trend < 0.05). Each 1 mu mol/L increment in maternal zinc level was associated with a 9.553 g (95 % CI: -16.370, -2.735) decrease in birth weight. A significant interaction between SOD2 promoter methylation and maternal serum zinc in relation to birth weight was observed in the AG+GG group (P-interaction < 0.05). Newborns carrying AA genotype were more sensitive to maternal serum zinc in the lower SOD2 group (Pinteraction < 0.05). Conclusions: Maternal serum zinc was inversely associated with birth weight, and this association was modified by neonatal SOD2 polymorphism and promoter methylation. These findings suggest that SOD2 polymorphism and promoter methylation may influence the relationship between maternal zinc status and fetal growth.

The time courses (TC) and functional network connectivity (FNC) features extracted from functional magnetic resonance imaging (fMRI) data demonstrate substantial potential in the investigation of brain disorders. Nevertheless, prevailing diagnostic approaches for brain disorders utilizing fMRI predominantly rely on either single-feature analyses or simplistic feature concatenation strategies, thereby neglecting the complementary information inherent in the interplay between temporal dynamics and spatial connectivity patterns. To address this problem, we propose Cross-Feature Mutual Learning (CFML), a novel framework that enables collaborative learning between feature-specific models through mutual knowledge transfer during training. Unlike traditional data-level fusion or decision-level ensemble methods, CFML introduces three key innovations: (1) a cross-feature mutual learning paradigm where TC and FNC encoders collaboratively exchange knowledge to unlock complementary information, (2) an end-to-end framework that jointly optimizes feature extraction, cross-modal knowledge transfer, and classification without requiring manual fusion design, and (3) a contrastive learning-based feature exchange mechanism that maximizes mutual information between heterogeneous brain feature networks. Extensive experiments on large multi-site datasets demonstrate CFML's superior performance, achieving 85.1 % and 72.3 % accuracy for schizophrenia and autism classification respectively, outperforming 16 state-of-the-art methods by 3.0-9.2 %. Attention analysis highlights biologically relevant areas, confirming the framework's clinical value. In summary, CFML represents a significant advance in neuroimaging-based mental disorder diagnosis with its innovative mutual learning approach.

Understanding white matter (WM) functional connectivity is crucial for unraveling brain function and dysfunction. In this study, we present a novel WM intrinsic connectivity network (ICN) template derived from over 100,000 fMRI scans, identifying 97 robust WM ICNs using spatially constrained independent component analysis (scICA). This WM template, combined with a previously identified gray matter (GM) ICN template from the same dataset, was applied to analyze a resting-state fMRI (rs-fMRI) dataset from the Bipolar-Schizophrenia Network on Intermediate Phenotypes 2 (BSNIP2; 590 subjects) and a task-based fMRI dataset from the MIND Clinical Imaging Consortium (MCIC; 75 subjects). Our analysis highlights distinct spatial maps for WM and GM ICNs, with WM ICNs showing higher frequency profiles. Visually modular structure within WM ICNs and interactions between WM and GM modules were identified. Task-based fMRI revealed event-related BOLD signals in WM ICNs, particularly within the corticospinal tract, lateralized to finger movement. Notable differences in static functional network connectivity (sFNC) matrices were observed between controls (HC) and schizophrenia (SZ) subjects in both WM and GM networks. This open-source WM NeuroMark template and automated pipeline offer a powerful tool for advancing WM connectivity research across diverse datasets.

Background and Hypothesis Schizophrenia (SZ) is characterized by significant cognitive and behavioral disruptions. Neuroimaging techniques, particularly magnetic resonance imaging (MRI), have been widely utilized to investigate biomarkers of SZ, distinguish SZ from healthy conditions or other mental disorders, and explore biotypes within SZ or across SZ and other mental disorders, which aim to promote the accurate diagnosis of SZ. In China, research on SZ using MRI has grown considerably in recent years.Study Design The article reviews advanced neuroimaging and artificial intelligence (AI) methods using single-modal or multimodal MRI to reveal the mechanism of SZ and promote accurate diagnosis of SZ, with a particular emphasis on the achievements made by Chinese scholars around the past decade.Study Results Our article focuses on the methods for capturing subtle brain functional and structural properties from the high-dimensional MRI data, the multimodal fusion and feature selection methods for obtaining important and sparse neuroimaging features, the supervised statistical analysis and classification for distinguishing disorders, and the unsupervised clustering and semi-supervised learning methods for identifying neuroimage-based biotypes. Crucially, our article highlights the characteristics of each method and underscores the interconnections among various approaches regarding biomarker extraction and neuroimage-based diagnosis, which is beneficial not only for comprehending SZ but also for exploring other mental disorders.Conclusions We offer a valuable review of advanced neuroimage analysis and AI methods primarily focused on SZ research by Chinese scholars, aiming to promote the diagnosis, treatment, and prevention of SZ, as well as other mental disorders, both within China and internationally.