Identifying and validating subtypes within major psychiatric disorders based on frontal-posterior functional imbalance via deep learning

Neuroimage Analysis Methods and Artificial Intelligence Techniques for Reliable Biomarkers and Accurate Diagnosis of Schizophrenia: Achievements Made by Chinese Scholars Around the Past Decade

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.

Aberrant Hippocampal Development in Early-onset Mental Disorders and Promising Interventions: Evidence from a Translational Study

Early-onset mental disorders are associated with disrupted neurodevelopmental processes during adolescence. The methylazoxymethanol acetate (MAM) animal model, in which disruption in neurodevelopmental processes is induced, mimics the abnormal neurodevelopment associated with early-onset mental disorders from an etiological perspective. We conducted longitudinal structural magnetic resonance imaging (MRI) scans during childhood, adolescence, and adulthood in MAM rats to identify specific brain regions and critical windows for intervention. Then, the effect of repetitive transcranial magnetic stimulation (rTMS) intervention on the target brain region during the critical window was investigated. In addition, the efficacy of this intervention paradigm was tested in a group of adolescent patients with early-onset mental disorders (diagnosed with major depressive disorder or bipolar disorder) to evaluate its clinical translational potential. The results demonstrated that, compared to the control group, the MAM rats exhibited significantly lower striatal volume from childhood to adulthood (all P <0.001). In contrast, the volume of the hippocampus did not show significant differences during childhood (P >0.05) but was significantly lower than the control group from adolescence to adulthood (both P <0.001). Subsequently, rTMS was applied to the occipital cortex, which is anatomically connected to the hippocampus, in the MAM models during adolescence. The MAM-rTMS group showed a significant increase in hippocampal volume compared to the MAM-sham group (P <0.01), while the volume of the striatum remained unchanged (P >0.05). In the clinical trial, adolescents with early-onset mental disorders showed a significant increase in hippocampal volume after rTMS treatment compared to baseline (P <0.01), and these volumetric changes were associated with improvement in depressive symptoms (r = - 0.524, P = 0.018). These findings highlight the potential of targeting aberrant hippocampal development during adolescence as a viable intervention for early-onset mental disorders with neurodevelopmental etiology as well as the promise of rTMS as a therapeutic approach for mitigating aberrant neurodevelopmental processes and alleviating clinical symptoms.

Characterizing the distinct imaging phenotypes, clinical behavior, and genetic vulnerability of brain maturational subtypes in mood disorders

BACKGROUND

Mood disorders are characterized by great heterogeneity in clinical manifestation. Uncovering such heterogeneity using neuroimaging-based individual biomarkers, clinical behaviors, and genetic risks, might contribute to elucidating the etiology of these diseases and support precision medicine.

METHODS

We recruited 174 drug-naïve and drug-free patients with major depressive disorder and bipolar disorder, as well as 404 healthy controls. T1 MRI imaging data, clinical symptoms, and neurocognitive assessments, and genetics were obtained and analyzed. We applied regional gray matter volumes (GMV) and quantile normative modeling to create maturation curves, and then calculated individual deviations to identify subtypes within the patients using hierarchical clustering. We compared the between-subtype differences in GMV deviations, clinical behaviors, cell-specific transcriptomic associations, and polygenic risk scores. We also validated the GMV deviations based subtyping analysis in a replication cohort.

RESULTS

Two subtypes emerged: subtype 1, characterized by increased GMV deviations in the frontal cortex, cognitive impairment, a higher genetic risk for Alzheimer's disease, and transcriptionally associated with Alzheimer's disease pathways, oligodendrocytes, and endothelial cells; and subtype 2, displaying globally decreased GMV deviations, more severe depressive symptoms, increased genetic vulnerability to major depressive disorder and transcriptionally related to microglia and inhibitory neurons. The distinct patterns of GMV deviations in the frontal, cingulate, and primary motor cortices between subtypes were shown to be replicable.

CONCLUSIONS

Our current results provide vital links between MRI-derived phenotypes, spatial transcriptome, genetic vulnerability, and clinical manifestation, and uncover the heterogeneity of mood disorders in biological and behavioral terms.

Label-based meta-analysis of functional brain dysconnectivity across mood and psychotic disorders

BACKGROUND

Resting-state functional magnetic resonance imaging (rsfMRI) studies have revealed patterns of functional brain dysconnectivity in psychiatric disorders such as major depression disorder (MDD), bipolar disorder (BD) and schizophrenia (SZ). Although these disorders have been mostly studied in isolation, there is mounting evidence of shared neurobiological alterations across them.

METHODS

To uncover the nature of the relatedness between these psychiatric disorders, we conducted an innovative meta-analysis of dysconnectivity findings reported separately in MDD, BD and SZ. Rather than relying on a classical voxel level coordinate-based approach, our procedure extracted relevant neuroanatomical labels from text data and examined findings at the whole brain network level. Data were drawn from 428 rsfMRI studies investigating MDD (158 studies, 7429 patients/7414 controls), BD (81 studies, 3330 patients/4096 patients) and/or SZ (223 studies, 11,168 patients/11,754 controls). Permutation testing revealed commonalities and differences in hypoconnectivity and hyperconnectivity patterns across disorders.

RESULTS

Hypoconnectivity and hyperconnectivity patterns of higher-order cognitive (default-mode, fronto-parietal, cingulo-opercular) networks were similarly observed across the three disorders. By contrast, dysconnectivity of lower-order (somatomotor, visual, auditory) networks in some cases differed between disorders, notably dissociating SZ from BD and MDD.

CONCLUSIONS

Findings suggest that functional brain dysconnectivity of higher-order cognitive networks is largely transdiagnostic in nature while that of lower-order networks may best discriminate between mood and psychotic disorders, thus emphasizing the relevance of motor and sensory networks to psychiatric neuroscience.

BPI-GNN: Interpretable brain network-based psychiatric diagnosis and subtyping

Converging evidence increasingly suggests that psychiatric disorders, such as major depressive disorder (MDD) and autism spectrum disorder (ASD), are not unitary diseases, but rather heterogeneous syndromes that involve diverse, co-occurring symptoms and divergent responses to treatment. This clinical heterogeneity has hindered the progress of precision diagnosis and treatment effectiveness in psychiatric disorders. In this study, we propose BPI-GNN, a new interpretable graph neural network (GNN) framework for analyzing functional magnetic resonance images (fMRI), by leveraging the famed prototype learning. In addition, we introduce a novel generation process of prototype subgraph to discover essential edges of distinct prototypes and employ total correlation (TC) to ensure the independence of distinct prototype subgraph patterns. BPI-GNN can effectively discriminate psychiatric patients and healthy controls (HC), and identify biological meaningful subtypes of psychiatric disorders. We evaluate the performance of BPI-GNN against 11 popular brain network classification methods on three psychiatric datasets and observe that our BPI-GNN always achieves the highest diagnosis accuracy. More importantly, we examine differences in clinical symptom profiles and gene expression profiles among identified subtypes and observe that our identified brain-based subtypes have the clinical relevance. It also discovers the subtype biomarkers that align with current neuro-scientific knowledge.

Individual deviations from normative electroencephalographic connectivity predict antidepressant response

BACKGROUND

Antidepressant medications yield unsatisfactory treatment outcomes in patients with major depressive disorder (MDD) with modest advantages over the placebo, partly due to the elusive mechanisms of antidepressant responses and unexplained heterogeneity in patient's response to treatment. Here we develop a novel normative modeling framework to quantify individual deviations in psychopathological dimensions that offers a promising avenue for the personalized treatment for psychiatric disorders.

METHODS

We built a normative model with resting-state electroencephalography (EEG) connectivity data from healthy controls of three independent cohorts. We characterized the individual deviation of MDD patients from the healthy norms, based on which we trained sparse predictive models for treatment responses of MDD patients (102 sertraline-medicated and 119 placebo-medicated). Hamilton depression rating scale (HAMD-17) was assessed at both baseline and after the eight-week antidepressant treatment.

RESULTS

We successfully predicted treatment outcomes for patients receiving sertraline (r = 0.43, p < 0.001) and placebo (r = 0.33, p < 0.001). We also showed that the normative modeling framework successfully distinguished subclinical and diagnostic variabilities among subjects. From the predictive models, we identified key connectivity signatures in resting-state EEG for antidepressant treatment, suggesting differences in neural circuit involvement between sertraline and placebo responses.

CONCLUSIONS

Our findings and highly generalizable framework advance the neurobiological understanding in the potential pathways of antidepressant responses, enabling more targeted and effective personalized MDD treatment.

TRIAL REGISTRATION

Establishing Moderators and Biosignatures of Antidepressant Response for Clinical Care for Depression (EMBARC), NCT#01407094.

A role for the cerebellum in motor-triggered alleviation of anxiety

Physical exercise is known to reduce anxiety, but the underlying brain mechanisms remain unclear. Here, we explore a hypothalamo-cerebello-amygdalar circuit that may mediate motor-dependent alleviation of anxiety. This three-neuron loop, in which the cerebellar dentate nucleus takes center stage, bridges the motor system with the emotional system. Subjecting animals to a constant rotarod engages glutamatergic cerebellar dentate neurons that drive PKCδ+ amygdalar neurons to elicit an anxiolytic effect. Moreover, challenging animals on an accelerated rather than a constant rotarod engages hypothalamic neurons that provide a superimposed anxiolytic effect via an orexinergic projection to the dentate neurons that activate the amygdala. Our findings reveal a cerebello-limbic pathway that may contribute to motor-triggered alleviation of anxiety and that may be optimally exploited during challenging physical exercise.

More reliable biomarkers and more accurate prediction for mental disorders using a label-noise filtering-based dimensional prediction method

The integration of neuroimaging with artificial intelligence is crucial for advancing the diagnosis of mental disorders. However, challenges arise from incomplete matching between diagnostic labels and neuroimaging. Here, we propose a label-noise filtering-based dimensional prediction (LAMP) method to identify reliable biomarkers and achieve accurate prediction for mental disorders. Our method proposes to utilize a label-noise filtering model to automatically filter out unclear cases from a neuroimaging perspective, and then the typical subjects whose diagnostic labels align with neuroimaging measures are used to construct a dimensional prediction model to score independent subjects. Using fMRI data of schizophrenia patients and healthy controls (n = 1,245), our method yields consistent scores to independent subjects, leading to more distinguishable relabeled groups with an enhanced classification accuracy of 31.89%. Additionally, it enables the exploration of stable abnormalities in schizophrenia. In summary, our LAMP method facilitates the identification of reliable biomarkers and accurate diagnosis of mental disorders using neuroimages.

Visual cortex repetitive transcranial magnetic stimulation (rTMS) reversing neurodevelopmental impairments in adolescents with major psychiatric disorders (MPDs): A cross-species translational study

AIMS

Neurodevelopmental impairments are closely linked to the basis of adolescent major psychiatric disorders (MPDs). The visual cortex can regulate neuroplasticity throughout the brain during critical periods of neurodevelopment, which may provide a promising target for neuromodulation therapy. This cross-species translational study examined the effects of visual cortex repetitive transcranial magnetic stimulation (rTMS) on neurodevelopmental impairments in MPDs.

METHODS

Visual cortex rTMS was performed in both adolescent methylazoxymethanol acetate (MAM) rats and patients with MPDs. Functional magnetic resonance imaging (fMRI) and brain tissue proteomic data in rats and fMRI and clinical symptom data in patients were analyzed.

RESULTS

The regional homogeneity (ReHo) analysis of fMRI data revealed an increase in the frontal cortex and a decrease in the posterior cortex in the MAM rats, representing the abnormal neurodevelopmental pattern in MPDs. In regard to the effects of rTMS, similar neuroimaging changes, particularly reduced frontal ReHo, were found both in MAM rats and adolescent patients, suggesting that rTMS may reverse the abnormal neurodevelopmental pattern. Proteomic analysis revealed that rTMS modulated frontal synapse-associated proteins, which may be the underpinnings of rTMS efficacy. Furthermore, a positive relationship was observed between frontal ReHo and clinical symptoms after rTMS in patients.

CONCLUSION

Visual cortex rTMS was proven to be an effective treatment for adolescent MPDs, and the underlying neural and molecular mechanisms were uncovered. Our study provides translational evidence for therapeutics targeting the neurodevelopmental factor in MPDs.

Transcriptomic Similarity Informs Neuromorphic Deviations in Depression Biotypes

BACKGROUND

Major depressive disorder (MDD) is complicated by population heterogeneity, motivating the investigation of biotypes through imaging-derived phenotypes. However, neuromorphic heterogeneity in MDD remains unclear, and how the correlated gene expression (CGE) connectome constrains these neuromorphic anomalies in MDD biotypes has not yet been studied.

METHODS

Here, we related cortical thickness deviations in MDD biotypes to a pattern of CGE connectome. Cortical thickness was estimated from 3-dimensional T1-weighted magnetic resonance images in 2 independent cohorts (discovery cohort: N = 425; replication cohort: N = 217). The transcriptional activity was measured according to Allen Human Brain Atlas. A density peak-based clustering algorithm was used to identify MDD biotypes.

RESULTS

We found that patients with MDD were clustered into 2 replicated biotypes based on single-patient regional deviations from healthy control participants across 2 datasets. Biotype 1 mainly exhibited cortical thinning across the brain, whereas biotype 2 mainly showed cortical thickening in the brain. Using brainwide gene expression data, we found that deviations of transcriptionally connected neighbors predicted regional deviation for both biotypes. Furthermore, putative CGE-informed epicenters of biotype 1 were concentrated on the cognitive control circuit, whereas biotype 2 epicenters were located in the social perception circuit. The patterns of epicenter likelihood were separately associated with depression- and anxiety-response maps, suggesting that epicenters of MDD biotypes may be associated with clinical efficacies.

CONCLUSIONS

Our findings linked the CGE connectome and neuromorphic deviations to identify distinct epicenters in MDD biotypes, providing insight into how microscale gene expressions informed MDD biotypes.

Microglia dysfunction drives disrupted hippocampal amplitude of low frequency after acute kidney injury

AIMS

Acute kidney injury (AKI) has been associated with a variety of neurological problems, while the neurobiological mechanism remains unclear. In the present study, we utilized resting-state functional magnetic resonance imaging (rs-fMRI) to detect brain injury at an early stage and investigated the impact of microglia on the neuropathological mechanism of AKI.

METHODS

Rs-fMRI data were collected from AKI rats and the control group with a 9.4-Tesla scanner at 24, 48, and 72 h post administration of contrast medium or saline. The amplitude of low-frequency fluctuations (ALFF) was then compared across the groups at each time course. Additionally, flow cytometry and SMART-seq2 were employed to evaluate microglia. Furthermore, pathological staining and Western blot were used to analyze the samples.

RESULTS

MRI results revealed that AKI led to a decreased ALFF in the hippocampus, particularly in the 48 h and 72 h groups. Additionally, western blot suggested that AKI-induced the neuronal apoptosis at 48 h and 72 h. Flow cytometry and confocal microscopy images demonstrated that AKI activated the aggregation of microglia into neurons at 24 h, with a strong upregulation of M1 polarization at 48 h and peaking at 72 h, accompanying with the release of proinflammatory cytokines. The ALFF value was strongly correlated with the proportion of microglia (|r| > 0.80, p < 0.001).

CONCLUSIONS

Our study demonstrated that microglia aggregation and inflammatory factor upregulation are significant mechanisms of AKI-induced neuronal apoptosis. We used fMRI to detect the alterations in hippocampal function, which may provide a noninvasive method for the early detection of brain injury after AKI.

Dimensional Neuroimaging Endophenotypes: Neurobiological Representations of Disease Heterogeneity Through Machine Learning

Machine learning has been increasingly used to obtain individualized neuroimaging signatures for disease diagnosis, prognosis, and response to treatment in neuropsychiatric and neurodegenerative disorders. Therefore, it has contributed to a better understanding of disease heterogeneity by identifying disease subtypes that present significant differences in various brain phenotypic measures. In this review, we first present a systematic literature overview of studies using machine learning and multimodal MRI to unravel disease heterogeneity in various neuropsychiatric and neurodegenerative disorders, including Alzheimer's disease, schizophrenia, major depressive disorder, autism spectrum disorder, multiple sclerosis, as well as their potential in transdiagnostic settings. Subsequently, we summarize relevant machine learning methodologies and discuss an emerging paradigm which we call dimensional neuroimaging endophenotype (DNE). DNE dissects the neurobiological heterogeneity of neuropsychiatric and neurodegenerative disorders into a low-dimensional yet informative, quantitative brain phenotypic representation, serving as a robust intermediate phenotype (i.e., endophenotype) largely reflecting underlying genetics and etiology. Finally, we discuss the potential clinical implications of the current findings and envision future research avenues.

LATTE: Label-efficient incident phenotyping from longitudinal electronic health records

Electronic health record (EHR) data are increasingly used to support real-world evidence studies but are limited by the lack of precise timings of clinical events. Here, we propose a label-efficient incident phenotyping (LATTE) algorithm to accurately annotate the timing of clinical events from longitudinal EHR data. By leveraging the pre-trained semantic embeddings, LATTE selects predictive features and compresses their information into longitudinal visit embeddings through visit attention learning. LATTE models the sequential dependency between the target event and visit embeddings to derive the timings. To improve label efficiency, LATTE constructs longitudinal silver-standard labels from unlabeled patients to perform semi-supervised training. LATTE is evaluated on the onset of type 2 diabetes, heart failure, and relapses of multiple sclerosis. LATTE consistently achieves substantial improvements over benchmark methods while providing high prediction interpretability. The event timings are shown to help discover risk factors of heart failure among patients with rheumatoid arthritis.

Integrative omics analysis reveals epigenomic and transcriptomic signatures underlying brain structural deficits in major depressive disorder

Several lines of evidence support the involvement of transcriptomic and epigenetic mechanisms in the brain structural deficits of major depressive disorder (MDD) separately. However, research in these two areas has remained isolated. In this study, we proposed an integrative strategy that combined neuroimaging, brain-wide gene expression, and peripheral DNA methylation data to investigate the genetic basis of gray matter abnormalities in MDD. The MRI T1-weighted images and Illumina 850 K DNA methylation microarrays were obtained from 269 patients and 416 healthy controls, and brain-wide transcriptomic data were collected from Allen Human Brain Atlas. The between-group differences in gray matter volume (GMV) and differentially methylated CpG positions (DMPs) were examined. The genes with their expression patterns spatially related to GMV changes and genes with DMPs were overlapped and selected. Using principal component regression, the associations between DMPs in overlapped genes and GMV across individual patients were investigated, and the region-specific correlations between methylation status and gene expression were examined. We found significant associations between the decreased GMV and DMPs methylation status in the anterior cingulate cortex, inferior frontal cortex, and fusiform face cortex regions. These DMPs genes were primarily enriched in the neurodevelopmental and synaptic transmission process. There was a significant negative correlation between DNA methylation and gene expression in genes associated with GMV changes of the frontal cortex in MDD. Our findings suggest that GMV abnormalities in MDD may have a transcriptomic and epigenetic basis. This imaging-transcriptomic-epigenetic integrative analysis provides spatial and biological links between cortical morphological deficits and peripheral epigenetic signatures in MDD.

A neuroimaging-based precision medicine framework for depression

BACKGROUND

Symptom-based diagnostic criteria of depression leads to notorious heterogeneity and subjectivity.

METHODS

The study was conducted in two stages at two sites: development of a neuroimaging-based subtyping and precise repetitive transcranial magnetic stimulation (rTMS) strategy for depression at Center 1 and its clinical application at Center 2. Center 1 identified depression subtypes and subtype-specific rTMS targets based on amplitude of low frequency fluctuation (ALFF) in a cohort of 238 major depressive disorder patients and 66 healthy controls (HC). Subtypes were identified using a Gaussian Mixture Model, and subtype-specific rTMS targets were selected based on dominant brain regions prominently differentiating depression subtypes from HC. Subsequently, one classifier was employed and 72 hospitalized, depressed youths at Center 2 received two-week precise rTMS. MRI and clinical assessments were obtained at baseline, midpoint, and treatment completion for evaluation.

RESULTS

Two neuroimaging subtypes of depression, archetypal and atypical depression, were identified based on distinct frontal-posterior functional imbalance patterns as measured by ALFF. The dorsomedial prefrontal cortex was identified as the rTMS target for archetypal depression, and the occipital cortex for atypical depression. Following precise rTMS, ALFF alterations were normalized in both archetypal and atypical depressed youths, corresponding with symptom response of 90.00% in archetypal depression and 70.73% in atypical depression.

CONCLUSIONS

A precision medicine framework for depression was developed based on objective neurobiomarkers and implemented with promising results, actualizing a subtyping-treatment-evaluation closed loop in depression. Future randomized controlled trials are warranted.

Mapping Neurophysiological Subtypes of Major Depressive Disorder Using Normative Models of the Functional Connectome

BACKGROUND

Major depressive disorder (MDD) is a highly heterogeneous disorder that typically emerges in adolescence and can occur throughout adulthood. Studies aimed at quantitatively uncovering the heterogeneity of individual functional connectome abnormalities in MDD and identifying reproducibly distinct neurophysiological MDD subtypes across the lifespan, which could provide promising insights for precise diagnosis and treatment prediction, are still lacking.

METHODS

Leveraging resting-state functional magnetic resonance imaging data from 1148 patients with MDD and 1079 healthy control participants (ages 11-93), we conducted the largest multisite analysis to date for neurophysiological MDD subtyping. First, we characterized typical lifespan trajectories of functional connectivity strength based on the normative model and quantitatively mapped the heterogeneous individual deviations among patients with MDD. Then, we identified neurobiological MDD subtypes using an unsupervised clustering algorithm and evaluated intersite reproducibility. Finally, we validated the subtype differences in baseline clinical variables and longitudinal treatment predictive capacity.

RESULTS

Our findings indicated great intersubject heterogeneity in the spatial distribution and severity of functional connectome deviations among patients with MDD, which inspired the identification of 2 reproducible neurophysiological subtypes. Subtype 1 showed severe deviations, with positive deviations in the default mode, limbic, and subcortical areas and negative deviations in the sensorimotor and attention areas. Subtype 2 showed a moderate but converse deviation pattern. More importantly, subtype differences were observed in depressive item scores and the predictive ability of baseline deviations for antidepressant treatment outcomes.

CONCLUSIONS

These findings shed light on our understanding of different neurobiological mechanisms underlying the clinical heterogeneity of MDD and are essential for developing personalized treatments for this disorder.

Metabolome subtyping reveals multi-omics characteristics and biological heterogeneity in major psychiatric disorders

Growing evidence suggests that major psychiatric disorders (MPDs) share common etiologies and pathological processes. However, the diagnosis is currently based on descriptive symptoms, which ignores the underlying pathogenesis and hinders the development of clinical treatments. This highlights the urgency of characterizing molecular biomarkers and establishing objective diagnoses of MPDs. Here, we collected untargeted metabolomics, proteomics and DNA methylation data of 327 patients with MPDs, 131 individuals with genetic high risk and 146 healthy controls to explore the multi-omics characteristics of MPDs. First, differential metabolites (DMs) were identified and we classified MPD patients into 3 subtypes based on DMs. The subtypes showed distinct metabolomics, proteomics and DNA methylation signatures. Specifically, one subtype showed dysregulation of complement and coagulation proteins, while the DNA methylation showed abnormalities in chemical synapses and autophagy. Integrative analysis in metabolic pathways identified the important roles of the citrate cycle, sphingolipid metabolism and amino acid metabolism. Finally, we constructed prediction models based on the metabolites and proteomics that successfully captured the risks of MPD patients. Our study established molecular subtypes of MPDs and elucidated their biological heterogeneity through a multi-omics investigation. These results facilitate the understanding of pathological mechanisms and promote the diagnosis and prevention of MPDs.

Dissociation between neuroanatomical and symptomatic subtypes in schizophrenia

BACKGROUND

Schizophrenia is a complex and heterogeneous syndrome with high clinical and biological stratification. Identifying distinctive subtypes can improve diagnostic accuracy and help precise therapy. A key challenge for schizophrenia subtyping is understanding the subtype-specific biological underpinnings of clinical heterogeneity. This study aimed to investigate if the machine learning (ML)-based neuroanatomical and symptomatic subtypes of schizophrenia are associated.

METHODS

A total of 314 schizophrenia patients and 257 healthy controls from four sites were recruited. Gray matter volume (GMV) and Positive and Negative Syndrome Scale (PANSS) scores were employed to recognize schizophrenia neuroanatomical and symptomatic subtypes using K-means and hierarchical methods, respectively.

RESULTS

Patients with ML-based neuroanatomical subtype-1 had focally increased GMV, and subtype-2 had widespread reduced GMV than the healthy controls based on either K-means or Hierarchical methods. In contrast, patients with symptomatic subtype-1 had severe PANSS scores than subtype-2. No differences in PANSS scores were shown between the two neuroanatomical subtypes; similarly, no GMV differences were found between the two symptomatic subtypes. Cohen's Kappa test further demonstrated an apparent dissociation between the ML-based neuroanatomical and symptomatic subtypes (P > 0.05). The dissociation patterns were validated in four independent sites with diverse disease progressions (chronic vs. first episodes) and ancestors (Chinese vs. Western).

CONCLUSIONS

These findings revealed a replicable dissociation between ML-based neuroanatomical and symptomatic subtypes of schizophrenia, which provides a new viewpoint toward understanding the heterogeneity of schizophrenia.

An overview of clustering methods with guidelines for application in mental health research

Cluster analyzes have been widely used in mental health research to decompose inter-individual heterogeneity by identifying more homogeneous subgroups of individuals. However, despite advances in new algorithms and increasing popularity, there is little guidance on model choice, analytical framework and reporting requirements. In this paper, we aimed to address this gap by introducing the philosophy, design, advantages/disadvantages and implementation of major algorithms that are particularly relevant in mental health research. Extensions of basic models, such as kernel methods, deep learning, semi-supervised clustering, and clustering ensembles are subsequently introduced. How to choose algorithms to address common issues as well as methods for pre-clustering data processing, clustering evaluation and validation are then discussed. Importantly, we also provide general guidance on clustering workflow and reporting requirements. To facilitate the implementation of different algorithms, we provide information on R functions and libraries.

Artificial intelligence in psychiatry research, diagnosis, and therapy

Psychiatric disorders are now responsible for the largest proportion of the global burden of disease, and even more challenges have been seen during the COVID-19 pandemic. Artificial intelligence (AI) is commonly used to facilitate the early detection of disease, understand disease progression, and discover new treatments in the fields of both physical and mental health. The present review provides a broad overview of AI methodology and its applications in data acquisition and processing, feature extraction and characterization, psychiatric disorder classification, potential biomarker detection, real-time monitoring, and interventions in psychiatric disorders. We also comprehensively summarize AI applications with regard to the early warning, diagnosis, prognosis, and treatment of specific psychiatric disorders, including depression, schizophrenia, autism spectrum disorder, attention-deficit/hyperactivity disorder, addiction, sleep disorders, and Alzheimer's disease. The advantages and disadvantages of AI in psychiatry are clarified. We foresee a new wave of research opportunities to facilitate and improve AI technology and its long-term implications in psychiatry during and after the COVID-19 era.

Volumes of hippocampal subfields suggest a continuum between schizophrenia, major depressive disorder and bipolar disorder

Objective

There is considerable debate as to whether the continuum of major psychiatric disorders exists and to what extent the boundaries extend. Converging evidence suggests that alterations in hippocampal volume are a common sign in psychiatric disorders; however, there is still no consensus on the nature and extent of hippocampal atrophy in schizophrenia (SZ), major depressive disorder (MDD) and bipolar disorder (BD). The aim of this study was to verify the continuum of SZ - BD - MDD at the level of hippocampal subfield volume and to compare the volume differences in hippocampal subfields in the continuum.

Methods

A total of 412 participants (204 SZ, 98 MDD, and 110 BD) underwent 3 T MRI scans, structured clinical interviews, and clinical scales. We segmented the hippocampal subfields with FreeSurfer 7.1.1 and compared subfields volumes across the three diagnostic groups by controlling for age, gender, education, and intracranial volumes.

Results

The results showed a gradual increase in hippocampal subfield volumes from SZ to MDD to BD. Significant volume differences in the total hippocampus and 13 of 26 hippocampal subfields, including CA1, CA3, CA4, GC-ML-DG, molecular layer and the whole hippocampus, bilaterally, and parasubiculum in the right hemisphere, were observed among diagnostic groups. Medication treatment had the most effect on subfields of MDD compared to SZ and BD. Subfield volumes were negatively correlated with illness duration of MDD. Positive correlations were found between subfield volumes and drug dose in SZ and MDD. There was no significant difference in laterality between diagnostic groups.

Conclusion

The pattern of hippocampal volume reduction in SZ, MDD and BD suggests that there may be a continuum of the three disorders at the hippocampal level. The hippocampus represents a phenotype that is distinct from traditional diagnostic strategies. Combined with illness duration and drug intervention, it may better reflect shared pathophysiology and mechanisms across psychiatric disorders.

Multi-scale convolutional recurrent neural network for psychiatric disorder identification in resting-state EEG

Background

Accurate classification based on affordable objective neuroimaging biomarkers are important steps toward designing individualized treatment.

Methods

In this work, we investigated a deep learning classification model, multi-scale convolutional recurrent neural network (MCRNN), to explore psychiatric disorder-related biomarkers by leveraging the spatiotemporal information of resting-state EEG (rsEEG) using a multiple psychiatric disorder database containing 327 individuals diagnosed with schizophrenia, bipolar, major depressive disorders, and healthy controls. All subjects were mapped to a shared low-dimensional subspace for intuitively interpreting the inter-relationship and separation of psychiatric disorders.

Results

Psychiatric disorders were identified using rsEEG with high accuracy ranged from 78.6 to 91.3% in patient vs. controls two-class classification, and 68.2% in four-class classification. The control-to-schizophrenia trajectory interpretated by the model was consistent with the disease severity in clinical observation.

Conclusion

The MsRNN demonstrated a capability in extracting discriminative rsEEG biomarkers for psychiatric disorder classification, indicating its potential to facilitate our understanding of psychiatric disorders and monitoring interventions.

Individual Deviations from Normative Electroencephalographic Connectivity Predict Antidepressant Response

Antidepressant medications yield unsatisfactory treatment outcomes in patients with major depressive disorder (MDD) with modest advantages over the placebo. This modest efficacy is partly due to the elusive mechanisms of antidepressant responses and unexplained heterogeneity in patient's response to treatment - the approved antidepressants only benefit a portion of patients, calling for personalized psychiatry based on individual-level prediction of treatment responses. Normative modeling, a framework that quantifies individual deviations in psychopathological dimensions, offers a promising avenue for the personalized treatment for psychiatric disorders. In this study, we built a normative model with resting-state electroencephalography (EEG) connectivity data from healthy controls of three independent cohorts. We characterized the individual deviation of MDD patients from the healthy norms, based on which we trained sparse predictive models for treatment responses of MDD patients. We successfully predicted treatment outcomes for patients receiving sertraline (r = 0.43, p < 0.001) and placebo (r = 0.33, p < 0.001). We also showed that the normative modeling framework successfully distinguished subclinical and diagnostic variabilities among subjects. From the predictive models, we identified key connectivity signatures in resting-state EEG for antidepressant treatment, suggesting differences in neural circuit involvement between treatment responses. Our findings and highly generalizable framework advance the neurobiological understanding in the potential pathways of antidepressant responses, enabling more targeted and effective MDD treatment.

Neurodevelopmental trajectories, polygenic risk, and lipometabolism in vulnerability and resilience to schizophrenia

BACKGROUND

Schizophrenia (SZ) arises from a complex interplay involving genetic and molecular factors. Early intervention of SZ hinges upon understanding its vulnerability and resiliency factors in study of SZ and genetic high risk for SZ (GHR).

METHODS

Herein, using integrative and multimodal strategies, we first performed a longitudinal study of neural function as measured by amplitude of low frequency function (ALFF) in 21 SZ, 26 GHR, and 39 healthy controls to characterize neurodevelopmental trajectories of SZ and GHR. Then, we examined the relationship between polygenic risk score for SZ (SZ-PRS), lipid metabolism, and ALFF in 78 SZ, and 75 GHR in cross-sectional design to understand its genetic and molecular substrates.

RESULTS

Across time, SZ and GHR diverge in ALFF alterations of the left medial orbital frontal cortex (MOF). At baseline, both SZ and GHR had increased left MOF ALFF compared to HC (P < 0.05). At follow-up, increased ALFF persisted in SZ, yet normalized in GHR. Further, membrane genes and lipid species for cell membranes predicted left MOF ALFF in SZ; whereas in GHR, fatty acids best predicted and were negatively correlated (r = -0.302, P < 0.05) with left MOF.

CONCLUSIONS

Our findings implicate divergence in ALFF alteration in left MOF between SZ and GHR with disease progression, reflecting vulnerability and resiliency to SZ. They also indicate different influences of membrane genes and lipid metabolism on left MOF ALFF in SZ and GHR, which have important implications for understanding mechanisms underlying vulnerability and resiliency in SZ and contribute to translational efforts for early intervention.

Latent class analysis of psychotic-affective disorders with data-driven plasma proteomics

Data-driven approaches to subtype transdiagnostic samples are important for understanding heterogeneity within disorders and overlap between disorders. Thus, this study was conducted to determine whether plasma proteomics-based clustering could subtype patients with transdiagnostic psychotic-affective disorder diagnoses. The study population included 504 patients with schizophrenia, bipolar disorder, and major depressive disorder and 160 healthy controls, aged 19 to 65 years. Multiple reaction monitoring was performed using plasma samples from each individual. Pathologic peptides were determined by linear regression between patients and healthy controls. Latent class analysis was conducted in patients after peptide values were stratified by sex and divided into tertile values. Significant demographic and clinical characteristics were determined for the latent clusters. The latent class analysis was repeated when healthy controls were included. Twelve peptides were significantly different between the patients and healthy controls after controlling for significant covariates. Latent class analysis based on these peptides after stratification by sex revealed two distinct classes of patients. The negative symptom factor of the Brief Psychiatric Rating Scale was significantly different between the classes (t = -2.070, p = 0.039). When healthy controls were included, two latent classes were identified, and the negative symptom factor of the Brief Psychiatric Rating Scale was still significant (t = -2.372, p = 0.018). In conclusion, negative symptoms should be considered a significant biological aspect for understanding the heterogeneity and overlap of psychotic-affective disorders.

BrainTGL: A dynamic graph representation learning model for brain network analysis

Modeling the dynamics characteristics in functional brain networks (FBNs) is important for understanding the functional mechanism of the human brain. However, the current works do not fully consider the potential complex spatial and temporal correlations in human brain. To solve this problem, we propose a temporal graph representation learning framework for brain networks (BrainTGL). The framework involves a temporal graph pooling for eliminating the noisy edges as well as data inconsistency, and a dual temporal graph learning for capturing the spatio-temporal features of the temporal graphs. The proposed method has been evaluated in both tasks of brain disease (ASD, MDD and BD) diagnosis/gender classification (classification task) and subtype identification (clustering task) on the four datasets: Human Connectome Project (HCP), Autism Brain Imaging Data Exchange (ABIDE), NMU-MDD and NMU-BD. A large improvement is achieved for the ASD diagnosis. Specifically, our model outperforms the GroupINN and ST-GCN by an average increase of 4.2% and 8.6% on accuracy, respectively, demonstrating its advantages in comparison to the state-of-the-art methods based on functional connectivity features or learned spatio-temporal features. The results demonstrate that learning the spatial-temporal brain network representation for modeling dynamics characteristics in FBNs can improve the model's performance on both disease diagnosis and subtype identification tasks for multiple disorders. Apart from performance, the improvements of computational efficiency and convergence speed reduce training costs.

Linking Personalized Brain Atrophy to Schizophrenia Network and Treatment Response

BACKGROUND AND HYPOTHESIS

Schizophrenia manifests with marked heterogeneity in both clinical presentation and underlying biology. Modeling individual differences within clinical cohorts is critical to translate knowledge reliably into clinical practice. We hypothesized that individualized brain atrophy in patients with schizophrenia may explain the heterogeneous outcomes of repetitive transcranial magnetic stimulation (rTMS).

STUDY DESIGN

The magnetic resonance imaging (MRI) data of 797 healthy subjects and 91 schizophrenia patients (between January 1, 2015, and December 31, 2020) were retrospectively selected from our hospital database. The healthy subjects were used to establish normative reference ranges for cortical thickness as a function of age and sex. Then, a schizophrenia patient's personalized atrophy map was computed as vertex-wise deviations from the normative model. Each patient's atrophy network was mapped using resting-state functional connectivity MRI from a subgroup of healthy subjects (n = 652). In total 52 of the 91 schizophrenia patients received rTMS in a randomized clinical trial (RCT). Their longitudinal symptom changes were adopted to test the clinical utility of the personalized atrophy map.

RESULTS

The personalized atrophy maps were highly heterogeneous across patients, but functionally converged to a putative schizophrenia network that comprised regions implicated by previous group-level findings. More importantly, retrospective analysis of rTMS-RCT data indicated that functional connectivity of the personalized atrophy maps with rTMS targets was significantly associated with the symptom outcomes of schizophrenia patients.

CONCLUSIONS

Normative modeling can aid in mapping the personalized atrophy network associated with treatment outcomes of patients with schizophrenia.

Recent advances in psychoradiology

Psychiatry, as a field, lacks objective markers for diagnosis, progression, treatment planning, and prognosis, in part due to difficulties studying the brain in vivo, and diagnoses are based on self-reported symptoms and observation of patient behavior and cognition. Rapid advances in brain imaging techniques allow clinical investigators to noninvasively quantify brain features at the structural, functional, and molecular levels. Psychoradiology is an emerging discipline at the intersection of psychiatry and radiology. Psychoradiology applies medical imaging technologies to psychiatry and promises not only to improve insight into structural and functional brain abnormalities in patients with psychiatric disorders but also to have potential clinical utility. We searched for representative studies related to recent advances in psychoradiology through May 1, 2022, and conducted a selective review of 165 references, including 75 research articles. We summarize the novel dynamic imaging processing methods to model brain networks and present imaging genetics studies that reveal the relationship between various neuroimaging endophenotypes and genetic markers in psychiatric disorders. Furthermore, we survey recent advances in psychoradiology, with a focus on future psychiatric diagnostic approaches with dimensional analysis and a shift from group-level to individualized analysis. Finally, we examine the application of machine learning in psychoradiology studies and the potential of a novel option for brain stimulation treatment based on psychoradiological findings in precision medicine. Here, we provide a summary of recent advances in psychoradiology research, and we hope this review will help guide the practice of psychoradiology in the scientific and clinical fields.

Modern views of machine learning for precision psychiatry

In light of the National Institute of Mental Health (NIMH)'s Research Domain Criteria (RDoC), the advent of functional neuroimaging, novel technologies and methods provide new opportunities to develop precise and personalized prognosis and diagnosis of mental disorders. Machine learning (ML) and artificial intelligence (AI) technologies are playing an increasingly critical role in the new era of precision psychiatry. Combining ML/AI with neuromodulation technologies can potentially provide explainable solutions in clinical practice and effective therapeutic treatment. Advanced wearable and mobile technologies also call for the new role of ML/AI for digital phenotyping in mobile mental health. In this review, we provide a comprehensive review of ML methodologies and applications by combining neuroimaging, neuromodulation, and advanced mobile technologies in psychiatry practice. We further review the role of ML in molecular phenotyping and cross-species biomarker identification in precision psychiatry. We also discuss explainable AI (XAI) and neuromodulation in a closed human-in-the-loop manner and highlight the ML potential in multi-media information extraction and multi-modal data fusion. Finally, we discuss conceptual and practical challenges in precision psychiatry and highlight ML opportunities in future research.

Shared and disease-sensitive dysfunction across bipolar and unipolar disorder during depressive episodes: a transdiagnostic study

Patients with depressive episodes (PDE), such as unipolar disorder (UD) and bipolar disorder (BD), are often defined as distinct diagnostic categories, but increasing converging evidence indicated shared etiologies and pathophysiological characteristics across different clinical diagnoses. We explored whether these transdiagnostic deficits are caused by the common neural substrates across diseases or disease-sensitive mechanisms, or a combination of both. In this study, we utilized a Bayesian model to decompose the resting-state brain activity into multiple hyper- and hypo-activity patterns (refer to as "factors"), so as to explore the shared and disease-sensitive alteration patterns in PDE. The model was constructed over a total of 259 patients (131 UD and 128 BD) with 100 healthy controls as the reference. The other 32 initial depressive episode BD (IDE-BD) patients who had symptoms of mania or hypomania during follow-up were taken as an independent set to estimate the factor composition using the established model for further analysis. We revealed three transdiagnostic alteration factors in PDE. Based on the distribution of factors and the tendency of factor composition at the group level, these factors were defined as BD sensitive factor, UD sensitive factor and shared basic alteration factor. We further found that the factor composition and the ROIs-based alteration degree (mainly involving in orbitofrontal gyrus and part of parietal lobe) were associated with the bipolar index in IDE-BD patients. Our findings contributed to understanding the core transdiagnostic shared and disease-sensitive alterations in PDE and to predicting the risk of emotional state transition in IDE-BD patients.

Higher benefit-risk ratio of COVID-19 vaccination in patients with schizophrenia and major depressive disorder versus patients with bipolar disorder when compared to controls

Patients with major psychiatric disorders (MPD) that include schizophrenia (SCH), bipolar disorder (BP), and major depressive disorder (MDD) are at increased risk for coronavirus disease 2019 (COVID-19). However, the safety and efficacy of COVID-19 vaccines in MPD patients have not been fully evaluated. This study aimed to investigate adverse events (AEs)/side effects and efficacy of COVID-19 vaccines in MPD patients. This retrospective study included 2034 patients with SCH, BP, or MDD who voluntarily received either BBIBP-CorV or Sinovac COVID-19 vaccines, and 2034 matched healthy controls. The incidence of AEs/side effects and the efficacy of COIVD-19 vaccinations among the two groups were compared. The risk ratio (RR) of side effects in patients with MPD was 0.60 (95% confidence interval [CI]: 0.53-0.68) after the first dose and 0.80 (95% CI: 0.65-0.99) following the second dose, suggesting a significantly lower risk in the MPD group versus healthy controls. The RRs of AEs did not differ between patients and controls. Notably, fully vaccinated patients exhibited a decreased risk of influenza with or without fever compared with controls (RR=0.38, 95% CI: 0.31-0.46; RR=0.23, 95% CI: 0.17-0.30; respectively). Further subgroup comparisons revealed a significantly lower risk of influenza with fever in MDD (RR=0.13, 95% CI: 0.08-0.21) and SCH (RR=0.24, 95% CI: 0.17-0.34) than BP (RR=0.85, 95% CI: 0.69-1.06) compared to controls. We conclude that the benefit-risk ratio of COVID-19 vaccination was more favorable in SCH or MDD versus BP when compared with controls. These data indicate that COVID-19 vaccines are safe and protective in patients with MPD from COVID-19.

Metabolic risk factors of cognitive impairment in young women with major psychiatric disorder

BACKGROUND

Cognitive performance improves clinical outcomes of patients with major psychiatric disorder (MPD), but is impaired by hyperglycemia. Psychotropic agents often induce metabolism syndrome (MetS). The identification of modifiable metabolic risk factors of cognitive impairment may enable targeted improvements of patient care.

OBJECTIVE

To investigate the relationship between MetS and cognitive impairment in young women with MPD, and to explore risk factors.

METHODS

We retrospectively studied women of 18-34 years of age receiving psychotropic medications for first-onset schizophrenia (SCH), bipolar disorder (BP), or major depressive disorder (MDD). Data were obtained at four time points: presentation but before psychotropic medication; 4-8 and 8-12 weeks of psychotropic therapy; and enrollment. MATRICS Consensus Cognitive Battery, (MCCB)-based Global Deficit Scores were used to assess cognitive impairment. Multiple logistic analysis was used to calculate risk factors. Multivariate models were used to investigate factors associated with cognitive impairment.

RESULTS

We evaluated 2,864 participants. Cognitive impairment was observed in 61.94% of study participants, and was most prevalent among patients with BP (69.38%). HbA1c within the 8-12 week-treatment interval was the most significant risk factor and highest in BP. Factors in SCH included pre-treatment waist circumference and elevated triglycerides during the 8-12 weeks treatment interval. Cumulative dosages of antipsychotics, antidepressants, and valproate were associated with cognitive impairment in all MPD subgroups, although lithium demonstrated a protect effect (all P < 0.001).

CONCLUSIONS

Cognitive impairment was associated with elevated HbA1c and cumulative medication dosages. Pre-treatment waist circumference and triglyceride level at 8-12 weeks were risk factors in SCH. Monitoring these indices may inform treatment revisions to improve clinical outcomes.

Shared and distinct functional connectivity of hippocampal subregions in schizophrenia, bipolar disorder, and major depressive disorder

Schizophrenia (SZ), bipolar disorder (BD), and major depressive disorder (MDD) share etiological and pathophysiological characteristics. Although neuroimaging studies have reported hippocampal alterations in SZ, BD, and MDD, little is known about how different hippocampal subregions are affected in these conditions because such subregions, namely, the cornu ammonis (CA), dentate gyrus (DG), and subiculum (SUB), have different structural foundations and perform different functions. Here, we hypothesize that different hippocampal subregions may reflect some intrinsic features among the major psychiatric disorders, such as SZ, BD, and MDD. By investigating resting functional connectivity (FC) of each hippocampal subregion among 117 SZ, 103 BD, 96 MDD, and 159 healthy controls, we found similarly and distinctly changed FC of hippocampal subregions in the three disorders. The abnormal functions of middle frontal gyrus might be the core feature of the psychopathological mechanisms of SZ, BD, and MDD. Anterior cingulate cortex and inferior orbital frontal gyrus might be the shared abnormalities of SZ and BD, and inferior orbital frontal gyrus is also positively correlated with depression and anxiety symptoms in SZ and BD. Caudate might be the unique feature of SZ and showed a positive correlation with the cognitive function in SZ. Middle temporal gyrus and supplemental motor area are the differentiating features of BD. Our study provides evidence for the different functions of different hippocampal subregions in psychiatric pathology.

The colors of our brain: an integrated approach for dimensionality reduction and explainability in fMRI through color coding (i-ECO)

Several systematic reviews have highlighted the role of multiple sources in the investigation of psychiatric illness. For what concerns fMRI, the focus of recent literature preferentially lies on three lines of research, namely: functional connectivity, network analysis and spectral analysis. Data was gathered from the UCLA Consortium for Neuropsychiatric Phenomics. The sample was composed by 130 neurotypicals, 50 participants diagnosed with Schizophrenia, 49 with Bipolar disorder and 43 with ADHD. Single fMRI scans were reduced in their dimensionality by a novel method (i-ECO) averaging results per Region of Interest and through an additive color method (RGB): local connectivity values (Regional Homogeneity), network centrality measures (Eigenvector Centrality), spectral dimensions (fractional Amplitude of Low-Frequency Fluctuations). Average images per diagnostic group were plotted and described. The discriminative power of this novel method for visualizing and analyzing fMRI results in an integrative manner was explored through the usage of convolutional neural networks. The new methodology of i-ECO showed between-groups differences that could be easily appreciated by the human eye. The precision-recall Area Under the Curve (PR-AUC) of our models was > 84.5% for each diagnostic group as evaluated on the test-set – 80/20 split. In conclusion, this study provides evidence for an integrative and easy-to-understand approach in the analysis and visualization of fMRI results. A high discriminative power for psychiatric conditions was reached. This proof-of-work study may serve to investigate further developments over more extensive datasets covering a wider range of psychiatric diagnoses.