Schizophrenia Research

Backgroundspeech carries cues to variation in mental state in schizophrenia spectrum disorders/psychotic disorders, typically indexed with clinician-rated scales such as the PANSS. Progress in the automation of speech-based symptom modelling has been constrained by data scale and the underrepresentation of low-resource languages. In this study, we aggregate multi-center recordings to assemble a large corpus and assess symptom-prediction models at scale, to enable more objective and efficient assessments and the early detection of relapse-related signals from speech. MethodsWe compiled data from 453 patients with schizophrenia spectrum disorders, recruited from ten global sites, and clipped their speech recordings into 6,664 segments. Across three feature sets, acoustic-prosodic profile, pretrained multilingual embeddings, and their concatenation, we compared 16 algorithms to predict eight relapse-related PANSS items, including three positive (P1, P2, P3), three negative (N1, N4, N6), and two general (G5, G9) items, on speaker-disjoint splits (80% train, 10% test, and 10% validation). Performance was assessed by root-mean-squared-error (RMSE) at both segment and participant (median aggregation) levels. Best model per item underwent bias checks for age, sex, education, and symptom severity. OutcomesBest-performing models predicted symptoms with prediction errors of 1{middle dot}5 PANSS points or lower: P1 1{middle dot}494/1{middle dot}527, P2 1{middle dot}318/1{middle dot}107, P3 1{middle dot}407/1{middle dot}542, N1 1{middle dot}029/1{middle dot}030, N4 1{middle dot}452/1{middle dot}430, N6 0{middle dot}860/0{middle dot}855, G5 0{middle dot}850/0{middle dot}882, G9 1{middle dot}213/1{middle dot}282 (segment/participant). Performance of the pretrained multilingual embeddings surpassed acoustic-prosodic features and their concatenation. Results were comparable in low-resource languages (e.g., Czech). We found no bias by age, sex, or education, aside from reduced N4 accuracy in males; but performance degraded with higher symptom severity. InterpretationSpeech can support automatic assessment of schizophrenia symptoms using pretrained multilingual embeddings, even without the use of transcripts. Such models show promise as clinically meaningful, efficient, and low-burden tools for real-time monitoring of symptom trajectories. FundingEU Horizon research and innovation programme. Research in contextO_ST_ABSEvidence before this studyC_ST_ABSAutomatic assessment of disease severity is a key issue in schizophrenia research, for which spontaneous speech offers a cost-effective, automatable solution. To evaluate existing evidence for speech-based symptom assessment, two reviewers (RHe, MK) searched PubMed, IEEE Xplore, arXiv, bioRxiv, and medRxiv for publications from inception to Aug 25, 2025, using the terms: ("symptom" OR "PANSS" OR "Positive and Negative Syndrome Scale") AND ("psychosis" OR "schizophrenia") AND ("language" OR "speech" OR "spontaneous speech") AND ("prediction" OR "machine learning" OR "deep learning" OR "algorithm" OR "neural network" OR "AI" OR "artificial intelligence"). Fourteen studies on symptom-level modelling were identified. Ten studies dichotomized clinical scores (e.g., PANSS) into low vs high for classification: five used conventional ML (e.g., random forests) and five used neural networks, with F1 scores ranging from 0{middle dot}60-0{middle dot}85. The remaining four studies, and two of the ten studies as mentioned above, modelled raw scores directly as regression tasks. Two relied solely on conventional regressors and the rest used neural networks, with errors from 0{middle dot}487 for single items (scale 1-7) to 8{middle dot}04 for summed scores (scale 18-126). All studies used free speech for elicitation, except one study, which used a reading task. Three studies incorporated additional tasks, such as picture description and immediate recall. None were multilingual: nine were in English, three in Chinese, one in Swiss German, and one in Brazilian Portuguese. Features spanned a wide range, including acoustic-prosodic profiles, morpho-syntactic structure, semantic organization, pragmatics (including sentiments), and even visual features capturing movement during talking. Representations from pretrained language models were also widely employed. Sample sizes (counting patients with schizophrenia) were generally small: eleven studies enrolled <50 patients, one had 65, and only two exceeded 100 patients. Some increased their effective sample size via multiple recordings per patient or by adding healthy controls and/or patients with other psychiatric disorders (e.g., depression). Added value of this studyTo our knowledge, this is the first multilingual, speech-based study modelling schizophrenia symptom severity with machine learning approach, and it includes the largest cohort of patients with schizophrenia to date. We further increased effective sample size by using diverse elicitation tasks and segmenting recordings into clips. This multilingual corpus empowers the usage of complex models and supports transfer learning from high-resource languages (e.g., English) to low-resource ones (e.g., Czech). For each of eight selected relapse-related PANSS items, the best audio-only models achieved RMSE < 1{middle dot}5, underscoring clinical relevance. We assessed potential biases: no effects were found for age, sex, or education (except poorer N4 performance in males), though performance declined at higher symptom severity. Trained models are released for use. Implications of all the available evidenceWe show that speech is a powerful signal for automatic assessment of schizophrenia symptom severity and holds promise for relapse prediction, even without transcripts. The approach readily extends to incorporate textual features (from manual or automatic transcripts) and more advanced models. Prospective studies with repeated recordings across relapse episodes are needed to validate the utility of our models on relapse prediction, for the sake of supporting precision psychiatry while reducing clinician burden.

Understanding factors that predict the course of schizophrenia remains essential for improving long-term clinical management. Rate and severity of symptom exacerbations vary widely across individuals, and although prior studies have examined potential predictors, findings have been inconsistent and often limited by small samples, infrequent assessments, and non-standardized measures. Using data from phase 1 of the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE), which includes a large cohort with monthly standardized evaluations, this study investigated whether baseline negative symptom severity predicts risk of symptom exacerbation over time. Participants were 1139 adults aged 18-65 years meeting DSM-IV criteria for schizophrenia. Symptoms worsening or exacerbation was defined as a [&ge;]12-point increase from baseline on the PANSS total score. Cox regression survival models examined the association between baseline PANSS negative symptom tertiles and time to exacerbation, adjusting for age, sex, PANSS positive and general psychopathology subscales, and CGI-Severity. Overall, 25.5% of participants experienced exacerbation over a 18-month period of follow-up. Survival curves demonstrated significant separation across negative symptom tertiles (p=0.047), with higher baseline negative symptoms associated with longer time to exacerbation. Compared with the lowest tertile, medium and high negative symptom groups showed reduced exacerbation risk (HR=0.73 and HR=0.69, respectively; both p=0.03). Findings indicate that greater baseline negative symptom severity is associated with a lower likelihood of short-term symptom worsening, suggesting a relatively stable illness course among individuals with more severe negative symptoms. These results have implications for prognosis and treatment planning, while underscoring the persistent functional burden imposed by negative symptoms despite lower exacerbation risk.

ObjectiveCharacterization of psychosis typically relies on cognitive and behavioral assessments. This study suggests the use of feature-specific sensory experiments to detect subtle perceptual alterations in early psychosis. MethodsPatients (N=120) diagnosed with schizophrenia (SCHZ, N=45), bipolar disorder (BIP, N=36), or first-episode psychosis (FEP, N=39), recruited from public mental health facilities in Brazil, were compared with age-matched healthy controls (HCSCHZ, HCBIP, and HCFEP; pooled from N=94). Independent psychophysical measurements were obtained within each group. The Pictorial-Size-Test (PST) assessed pictorial size perception. Sound-Appreciation-Test (SAT) assessed auditory discomfort. ResultsSCHZ circled larger perceived sizes than HCSCHZ (power=95%, d=0.63, p<0.0001), FEP circled larger perceived sizes than HCFEP (power=99%, d=2.86, p<0.0001), but BIP did not perceive larger sizes than HCBIP in PST. SCHZ reported higher levels of discomfort than HCSCHZ (power=99%, d=1.29, p<0.0005), BIP reported higher levels of discomfort than HCBIP (power=99%, d=2.73, p<0.0001) and FEP reported higher levels of discomfort than HCFEP (power=99%, d=1.46, p<0.0003) on SAT. ConclusionsThe findings suggest that low-cost psychophysical measurements can provide information about sensory alterations in early psychosis revealing dissimilar patterns between schizophrenia and bipolar disorder. Such patterns are not readily perceived by physician-patient interaction but may add to overall clinical judgement.

BackgroundPoor patient retention and family engagement compromise the effectiveness of coordinated specialty care (CSC) for first-episode psychosis (FEP). This mixed methods study aimed to identify program-level characteristics (CSC fidelity and engagement strategies) associated with patient retention and family engagement in Massachusetts CSC programs. MethodsPrimary outcomes were rates of patient retention and family engagement ([&ge;]1 evidence-based family intervention session), based on CSC program census (October 2022 - September 2023). Quantitative analyses explored program characteristics (EPINET Program-Level Core Assessment Battery) and fidelity ratings (Massachusetts Psychosis Fidelity Scale) as predictors using t-tests or univariate linear regressions. Thematic analysis of program interviews compared patient and family engagement strategies employed by high versus low performing programs. ResultsAcross nine programs, mean patient retention was 86% (range: 58-97%) and family engagement was 40% (range: 12-100%). Higher fidelity to evidence-based services (e.g., individual therapy, family intervention, and supported education/employment) was significantly associated with both outcomes (p<.05; R2 range: .51-.72). Mixed-methods analysis showed that high performing programs used case management-related supports to meet service users practical needs. Factors associated with higher patient retention included having comprehensive intake assessments, provider visits during hospitalization, and periodic treatment reviews. Programs that conducted benefits counseling and proactively recommended family services as standard care had higher family engagement. ConclusionsHigher fidelity CSC programs had better patient retention and family engagement. Case management-related supports addressed treatment barriers. Strategies designed to strengthen therapeutic alliance and goal alignment may promote patient engagement, while family engagement may benefit from proactive recommendation of family intervention.

ObjectiveDespite recommendations that young people at clinical high risk (CHR) for psychosis receive stepped treatment, few programs have published details of their clinical models or outcomes. This study describes the preliminary effectiveness of a risk calculator-informed stepped care model used at the Yale PRIME Clinic, a specialized outpatient clinic for young people at CHR. MethodsSeventy-one individuals (ages 12-25) at CHR enrolled in Yales PRIME Clinic during the first four years of the treatment program. Participants completed clinical assessments at six timepoints over two years of treatment within a care model informed by an empirically grounded psychosis risk calculator. Linear mixed-effect models were fit to examine changes in clinical symptoms over time, and sensitivity analyses evaluated differences in clinical trajectories between completers and non-completers. ResultsIndividuals engaged in treatment demonstrated significant and sustained improvements in positive, negative, general, disorganized, and depressive symptoms. Improvements in positive symptoms emerged by 6 months and continued to improve across most subsequent timepoints (6, 12, and 24 months). Pattern mixture analyses suggested that clinical trajectories did not significantly differ between completers and non-completers, though non-completers possessed more heterogeneous trajectories. ConclusionsA stepped care model informed by individualized risk calculator scores was feasible for delivery in a specialized outpatient setting, and was associated with broad symptom improvement for young people at CHR. Further controlled studies with blinded raters are needed to further confirm the efficacy of stepped care models and isolate the active components of treatment. HighlightsO_LIParticipants at clinical high risk for psychosis experienced significant reductions in attenuated psychotic symptoms and improvements in mood while enrolled in a risk-calculator-informed stepped care treatment model. C_LIO_LIParticipants who disengaged from treatment did not have significantly different clinical trajectories than those who remained in care. C_LIO_LIThe results suggest preliminary evidence for the feasibility of implementing a risk-calculator-informed stepped care model. C_LI

Background and HypothesisClozapine is the only antipsychotic with protective effects against suicide in schizophrenia (SCZ). Newer third-generation antipsychotics (TGA) have better tolerability and modulate serotonin, dopamine, and N-methyl-d-aspartate neurotransmission pathways implicated in suicide. We aimed to investigate the effects of TGAs on suicide in SCZ. MethodsWe searched seven databases up to December 2023 for SCZ studies that reported suicide data. The primary outcome was suicide deaths and attempts; suicidal ideation was added as a secondary outcome. Random effects meta-analyses quantified suicide risk in randomized controlled trials (RCT) while single proportion meta-analyses assessed longitudinal suicide risk in open label extension trials (OLE). For RCTs, sensitivity analyses were conducted and subgroup analyses explored the impact of dose, drug type, and comparator arm. Study ResultsTwenty articles were included; thirteen excluded higher suicide risk participants. Compared to placebo control, TGAs did not significantly change the risk of primary [RR = 0.65, p = 0.38] or secondary [RR = 0.63, p = 0.15] suicide outcomes. Subgroup and sensitivity analyses were not statistically significant. For OLEs, there was a significant increase in the incidence of primary [Ip = 0.004, p = 0.048] and secondary [Ip = 0.024, p = 0.0013] suicide outcomes, but there was marked study heterogeneity. ConclusionThere is no current trial evidence to show that TGAs significantly impact suicide outcomes in SCZ. The signal from OLEs should be interpreted cautiously due to heterogeneity and requires replication. An effective clozapine alternative is needed for suicide prevention in SCZ.

BackgroundAntipsychotics are central to relapse prevention in schizophrenia, but longer-term use is associated with adverse effects that often prompt dose reduction or discontinuation. Although relapse risk increases after discontinuation, the nature of relapse remains unclear. Specifically, it is uncertain whether relapse reflects re-emergence of underlying illness or pharmacological withdrawal. MethodsWe analysed longitudinal symptom data (Positive and Negative Syndrome Scale; PANSS) from 417 individuals with schizophrenia who experienced relapse post-stabilization in five randomized, double-blind, placebo-controlled discontinuation trials of oral and long-acting injectable (LAI) paliperidone. Latent class mixed modelling was used to identify distinct trajectories of symptom change preceding relapse. FindingsTwo latent classes of relapse were identified: rapid and delayed onset. Rapid relapse was associated with more severe positive, negative, and cognitive symptoms at relapse. The proportion of individuals experiencing rapid relapse did not differ between those randomized to placebo (treatment discontinuation) versus treatment continuation in either LAI (p=0.119) or oral trials (p=0.949). No consistent increase in withdrawal-like symptoms (e.g. anxiety, agitation, depression) was found in discontinuation compared to continuation groups. Across formulations, individuals with rapid relapse had significantly higher baseline PANSS scores than those with delayed relapse (p<0.001). InterpretationRelapse following antipsychotic discontinuation follows at least two distinct trajectories that are not specific to treatment withdrawal. The comparable proportions of rapid and delayed relapse trajectories between discontinuation and continuation groups, together with the absence of a distinct symptom profile at relapse, does not support pharmacological withdrawal as a common mechanism of relapse. Instead, higher baseline symptom severity in those who experience rapid relapse may reflect pre-existing vulnerability and/or trial-related measurement artefacts related to baseline symptom rating and trial inclusion criteria. This emphasizes the clinical importance of risk stratification and individual monitoring, and challenge the assumption that relapse risk can be meaningfully reduced through dose tapering strategies alone.

BackgroundAntipsychotic medications are recommended for managing schizophrenia spectrum disorders, yet their long-term effects on functional recovery remain unclear. Existing evidence is conflicting, often derived from between-subject comparisons vulnerable to confounding by indication. MethodsWe conducted a nationwide register-based cohort study of 65,630 individuals with incident schizophrenia spectrum disorders in Denmark (1998-2023). We modelled antipsychotic exposure against productive engagement (employment or education). We employed two distinct analytical approaches to separate causal from prognostic associations: (1) Within-subject stratified Cox models with time-varying covariates, where patients served as their own controls to eliminate time-invariant confounding; and (2) Fine-Gray competing risks models using a between-subject design with baseline exposure, accounting for mortality and emigration. FindingsOver 26.9 million person-weeks, the overall productive engagement rate was 48.2%. Integration of hospital pharmacy data revealed a 6.1% exposure misclassification bias in previous studies relying solely on community records. The primary within-subject analysis revealed significant temporal heterogeneity: medication use was associated with reduced engagement rates in the acute (0-2 years: HR 0.908) and consolidation phases (2-5 years: HR 0.946), but reversed to a positive association in the maintenance phase (5+ years: HR 1.019). In contrast, the between-subject Fine-Gray model yielded a null result (SHR 1.002, 95% CI 0.988-1.015), failing to detect these phase-specific dynamics. InterpretationWithin-subject modelling reveals that antipsychotic treatment involves a functional trade-off: it is associated with a transient reduction in engagement rates during the early consolidation phase (2-5 years), followed by stabilisation and potential benefit in the maintenance phase (5+ years). The null result in standard between-subject (Fine-Gray) analysis likely reflects residual confounding by indication and exposure misclassification, highlighting the necessity of within-person designs to unmask the true stage-specific impact of pharmacotherapy on vocational recovery. FundingNone directly for this study. Danmarks Nationalbank funded the research visit that facilitated this collaboration.

Schizophrenia is a severe neuropsychiatric disorder characterized by positive and negative symptoms and cognitive impairments. The present study aimed to investigate the potential interference of ambient noise on the performance of executive function (EF) tasks in individuals with schizophrenia. The sample consisted of 40 participants, divided equally into two groups: a group of individuals with schizophrenia (SchG) and a healthy control group without neuropsychiatric disorders (HC). All participants did three EF assessment instruments: Trail Making Test, Corsi Block Test, and Maze Test. The experimental design included a test-retest procedure with order counterbalancing: half of the sample began the assessment in the noise condition and the other half in the no-noise condition, to control for order and learning effects. The results indicate that ambient noise has a negative impact on the cognitive performance of individuals with schizophrenia. Specifically, the SchG group performed significantly worse on the Maze Test in the noise condition compared to the no-noise condition. These findings contribute to the understanding of the interactions between sensory and cognitive processes underlying the symptoms of schizophrenia. In addition to their theoretical potential, the results have practical implications, as they support the development of intervention strategies and ambiental adaptations that can improve the functionality and quality of life of people with the disorder.

Cognitive deficits in schizophrenia significantly hinder functional outcomes and often remain unresponsive to conventional treatments. While initial evidence suggested potential pro-cognitive effects of electrical brain stimulation in schizophrenia, recent meta-analyses have not supported these findings, warranting further investigation on intervention optimization. This sham-controlled crossover study explored cognitive and emotional effects of bilateral dorsolateral prefrontal cortex (DLPFC) anodal transcranial direct current stimulation (tDCS) and high-frequency transcranial random noise stimulation (tRNS) in schizophrenia. Thirty-six male patients with schizophrenia participated in a crossover trial, receiving three sessions (tDCS, tRNS, sham) in counterbalanced order with one-week intervals. tDCS and tRNS sessions involved 20-minute 2 mA anodal stimulation (tDCS) and 2 mA 100-640 Hz random noise stimulation targeting the left and right DLPFCs (F3-F4) with two extracephalic return electrodes. Executive functions (working memory, planning) were assessed during stimulation, and emotional changes were measured with the Positive and Negative Affect Schedule (PANAS) pre- and post-stimulation. Side effects and blinding efficacy were evaluated. Both bilateral tDCS and tRNS significantly improved executive functions (i.e., problem solving) compared to sham, with tRNS additionally enhancing working memory accuracy and strategy score. Both interventions increased positive affect and reduced negative affect after the intervention, with tRNS showing greater enhancement of positive emotions. Reduced negative affect correlated with better executive functions during tRNS. Side effects were minimal, and blinding was effective for the sham condition. Bilateral DLPFC anodal tDCS and high-frequency tRNS show promise as adjunctive treatments for schizophrenia, especially for cognitive deficits, with broader cognitive and emotional benefits observed with tRNS.

Cognitive heterogeneity is a core feature of schizophrenia (SCZ). Conventional approaches examine this heterogeneity using domain-specific scores, which may not fully reflect the underlying cognitive structure. In this study, a norm-anchored cognitive structural deviation (NCSD) framework was developed to examine such heterogeneity from a structure-informed perspective. The HC-derived latent cognitive structure (N-LCS) captured performance across the assessed tasks and remained stable under external validation in an independent cohort. Patients with SCZ showed greater global deviation from the N-LCS, along with altered loading directions of Wisconsin Card Sorting Test (WCST)-derived executive indicators which were consistent across robustness analyses, and altered correlation patterns among cognitive measures relative to HC. These features were quantified using three NCSD-derived indices: the cognitive normative deviation index (CNDI), loading pattern divergence (LPD), and correlation structure discrepancy (CSD). CNDI discriminated SCZ from HC with stable performance under cross-validation. LPD and CSD were associated with anxiety ratings, with LPD also showing a trend-level association with positive symptoms. Exploratory clustering identified a three-cluster solution with clear separation and moderate stability. Together, these findings show that cognitive heterogeneity in SCZ involves both global deviation from the N-LCS and structural alteration. NCSD provides a refined framework to characterize such heterogeneity and may inform precision psychiatry and functional recovery.

BackgroundPrediction of response to antipsychotic medications remains elusive, and a biomarker assisting in treatment selection would drastically improve prognosis. The 40 Hz auditory steady state response (ASSR) is an EEG biomarker, mirroring the GABA-glutamate signaling and the excitation/inhibition balance, consistently been reported to be impaired in schizophrenia, on, with inconsistent evidence of an association with specific symptoms. MethodsN=69 schizophrenia inpatients with an acute psychotic episode underwent an EEG recording to assess event related spectral perturbation (ERSP), intertrial phase coherence (ITC) and phase amplitude coupling (PAC) during the ASSR task, aimed to assess their relationship with response to antipsychotics and with positive, negative, disorganized, excited and depressive symptoms. Moreover, patients were compared with controls (N=36), to delineate schizophrenia acute phase ASSR dynamics. ResultsResponders to treatment showed a decreased 40 Hz ERSP in both the early (0-0.2s post-stimulus; P=0.0013; d=-0.936) and late (0-2-1.2s post-stimulus; P=0.0022; d=-0.932) time windows compared to non-responders. Using logistic regression and bootstrap optimism correction, ERSP classified the two groups with 70% accuracy. Responders but not non-responders showed a reduced ERSP compared to controls (P=0.0211; d=-0.558). Patients had reduced early ITPC (P=0.0001; d=-1.015) vs controls. responders compared to non-responders had increased PAC in the early (P=0.0215; d00.65) and in patients vs controls, in both the early (P=0.0002; d=0.57) and the late (P=0.0006; d=0.74) windows. No association emerged between ASSR metrics and symptoms severity. ConclusionsASSR is a candidate biomarker for antipsychotic treatment personalization. Only responders to treatment presented a significant gamma-band impairment, in line with previous literature on stabilized outpatients, but not non-responders, indicating that a distinct neurobiology could exist.

ObjectiveCognitive deficits are a leading cause of disability in schizophrenia and are linked to poor functional outcomes. There are no first line treatments for these deficits, and their neural basis is poorly understood. While schizophrenia is associated with widespread cognitive deficits, information processing speed is most profoundly impaired. Processing speed deficits have been associated with hyperconnectivity in the Default Mode Network (DMN). We therefore tested if modulating DMN connectivity with single or multiple sessions of transcranial magnetic stimulation (TMS) applied to an individualized DMN target would affect processing speed. MethodsIn the first study, 10 individuals with schizophrenia received single TMS sessions and underwent resting-state neuroimaging and processing speed assessment (Brief Assessment of Cognition in Schizophrenia digit symbol coding) acutely before and after each session. These sessions included excitatory (intermittent theta burst stimulation, iTBS); inhibitory (continuous theta burst stimulation, cTBS); and sham stimulation sessions. In the second study, 29 individuals (17 schizophrenia, 12 non-psychosis controls) received 5 accelerated sessions of cTBS with resting-state neuroimaging and processing speed assessment before and after the course of TMS sessions. ResultsIn the accelerated, multi-session DMN-targeted TMS trial, cTBS improved processing speed in the schizophrenia group (p=0.0124). In individuals with schizophrenia, reduction in DMN connectivity was linked to improvement in processing speed (p=0.021). These changes were dependent on age, where younger participants experienced greater processing speed improvements than older participants (p=0.006). ConclusionsIn sum, personalized network targeted TMS is a novel method for reducing cognitive impairment associated with schizophrenia.

Working memory (WM) impairments have been reported in different stages of psychosis but whether their neural correlates are shared or stage-specific is unknown. This meta-analysis examined WM-related brain activity across psychosis stages: familial and clinical high-risk for psychosis (at-risk stage), first-episode psychosis (early psychosis stage), and chronic schizophrenia (chronic psychosis stage). PubMed, Ovid, and Web of Science were searched up to July 2025 for functional magnetic resonance imaging (fMRI) studies comparing individuals in each stage and healthy controls during WM. Seed-based d-mapping assessed WM-related fMRI correlates at each stage. Significance was set at family-wise error-corrected p<.05. Forty-two studies were included: 7 in the at-risk stage, 5 in the early psychosis stage, and 30 in the chronic psychosis stage. In chronic psychosis, higher activation relative to controls was observed in the medial prefrontal cortex, rostral anterior cingulate, right insula and superior temporal gyrus, posterior cingulate cortex, left superior temporal and supramarginal gyri. Lower activation in chronic psychosis vs controls was found in the cerebellum, bilateral precuneus, middle temporal gyrus, and thalamus. The early psychosis stage was characterised by lower activation compared to controls in the dorsal anterior cingulate, bilateral caudate nuclei, and inferior frontal gyrus. No significant clusters emerged in the at-risk stage, or across stages. In combined early and chronic psychosis analyses, anterior cingulate cortex activation was positively associated with both antipsychotic dose and illness duration. These findings indicate that disruptions in WM circuitry may evolve after illness onset and may represent a potential biomarker of psychosis staging. HighlightsO_LIThis study revealed distinct brain activity patterns in early and chronic psychosis stages. C_LIO_LIAlterations in brain activity were more widespread in the chronic stage of psychosis. C_LIO_LIAntipsychotic dose and illness duration predicted anterior cingulate cortex activation. C_LIO_LIDistinct neural correlates of working memory may reflect illness progression. C_LI

Schizophrenia is a serious mental disorder characterized by enhanced sensory-perceptual alterations. We investigated face identity recognition in people with schizophrenia with the Facial Identity Recognition Structured Task (FIRST) develop at our laboratory. This was created with natural interference features (beard, makeup and mask). This task consists in six block-trails of six images for identity recognition. Forty three adult volunteers divided into two groups: a Health Control (HC) and a group of hospitalized patients with Schizophrenia (SchG) participated in the study. We measured the total number of correct answers as well as the average reaction time for each block. We observed significant losses in recognition of identity faces with interferences such as make up, beard and facial-mask.

BackgroundEEG microstates provide a window into rapid, large-scale brain network dynamics. Despite showing alterations in schizophrenia, evidence in first-episode schizophrenia spectrum psychosis (FESSP) is limited. We assessed whether microstate temporal and transition features could identify a multivariate signature of FESSP, and whether these dynamics can track symptom severity. MethodsResting-state EEG was analysed in 69 participants (FESSP n=41, mean age: 22.49 years; healthy controls n=28, mean age: 21.33 years). Twenty-eight microstate temporal and transition features were extracted across microstate classes (A-D). Group classification accuracy was assessed using a linear support vector machine with stratified cross-validation and permutation testing. Within the FESSP group, we further assessed associations between microstate features and clinical scores using the Brief Psychiatric Rating Scale (BPRS), Scale for the Assessment of Positive Symptoms (SAPS), and Scale for the Assessment of Negative Symptoms (SANS). ResultsMultivariate microstate features provided above-chance discrimination of FESSP from controls (balanced accuracy=0.644; AUC=0.688; p=0.030). However, when comparing individual features between groups, no feature survived multiple-comparison correction consistent with characterisation of FESSP via a distributed multivariate pattern across correlated features. Within the FESSP group, microstate dynamics were most strongly linked to negative symptoms, with higher SANS scores associated with shorter microstate D durations ({rho}=-0.507, pFDR=0.020) and higher occurrence of microstates A and B ({rho}=0.434-0.443, pFDR=0.042). BPRS-18 and SAPS showed no associations with any features. ConclusionsUsing EEG microstate temporal and transition features with multivariate classification, we identified a pattern that differentiated FESSP from controls and showed selective associations with negative symptom severity.

Background and HypothesisSchizophrenia is a disease characterized by various symptoms and has severe lifelong impacts on patients and their families. Despite various hypotheses and associated studies, the key mechanism in schizophrenia is not fully elucidated. In the present study, we focused on adropin, a peptide regulating energy metabolism, antioxidation, and neuroprotection. Study DesignIn both the group of healthy volunteers (HV) and the group of patients with some schizophrenia spectrum and other psychotic disorders (SZ), we evaluated adropin along with other variables such as anthropological factors, psychological well-being indicators, and laboratory test results. Study ResultsThe adropin levels in SZ were not significantly different from those in HV. Correlation analysis indicated five significant correlations beyond various natural correlations arising from fundamental proportional relationships and multifaceted psychological well-being indicators: (1) adropin versus right handgrip strength in the SZ group ({tau} = -0.82, P = 0.066); (2) adropin versus selenium in the total group ({tau} = 0.44, P = 0.053); (3) ferritin versus perceived stress in the total group ({tau} = -0.44, P = 0.053); (4) right versus left handgrip strength in the total group ({tau} = 0.70, P = 0.001) and in the SZ group ({tau} = 0.82, P = 0.075); and (5) selenium versus state anxiety in the total group ({tau} = 0.44, P = 0.053) and the SZ group ({tau} = 0.84, P = 0.066). ConclusionsThe present study provides a foundation for future studies and sheds light on the role of adropin in schizophrenia.

Auditory-verbal hallucinations (AVHs) are among the most disabling symptoms of schizophrenia and often persist despite the use of adequate antipsychotic treatment. Conventional low-frequency repetitive transcranial magnetic stimulation (rTMS) targeting the T3P3 scalp site has demonstrated limited efficacy, likely due to interindividual variability in AVH-related brain networks. In this multicenter, randomized, double-blind phase 3 trial, 70 patients with drug-resistant AVHs received active 1-Hz rTMS targeted either with an individualized fMRI-based symptom-capture procedure or by using conventional T3P3 localization. fMRI-guided rTMS yielded a greater reduction in Auditory Hallucination Rating Scale (AHRS) scores at one month (mean difference, -5.43; 95% CI, -8.92 to -1.94), and the effects were sustained at three and six months. The number-needed-to-treat for neuroguided rTMS was 3.5. Clinical response was associated with greater E-field overlap with AVH-related networks. These findings demonstrate that fMRI-guided neuronavigation increases rTMS efficacy, thus supporting its use to optimize the treatment of drug-resistant AVHs in schizophrenia.

ImportanceLarge language models (LLMs) have demonstrated diagnostic potential in several medical specialties, but their application to psychiatry - where diagnosis relies heavily on clinical judgment, narrative interpretation, and reasoning under uncertainty - remains insufficiently evaluated. ObjectiveTo evaluate diagnostic accuracy and clinician-judged reasoning quality of multiple large language models using psychiatric case vignettes. DesignMixed-methods evaluation study of diagnostic accuracy across four LLMs using 196 psychiatric case vignettes (135 published and 61 novel). Clinical reasoning quality was evaluated on a randomly selected subset of 30 vignettes using structured clinician ratings along two reasoning dimensions. The highest-performing model was illustratively compared with psychiatry trainees on the same subset. Diagnostic correctness for the full vignette set was assessed by a separate adjudicator LLM. SettingPublicly available model interfaces, December 2025. ParticipantsFive board-certified psychiatrists evaluated model-generated clinical reasoning. Two psychiatry residents served as the illustrative human comparison. Main Outcomes and MeasuresDiagnostic accuracy and clinician-rated clinical reasoning quality. Diagnostic accuracy was assessed using top-1 accuracy, top-5 accuracy, recall@5, and mean reciprocal rank based on ranked lists of five differential diagnoses per vignette. Clinical reasoning quality was assessed using two 5-point Likert scales adapted from the American Council of Graduate Medical Education Psychiatry Residency Milestones, evaluating data extraction and diagnostic reasoning. ResultsAcross 196 psychiatric case vignettes, Claude Opus 4.5 (Anthropic) achieved the highest diagnostic accuracy (top-1 accuracy, 0.638; top-5 accuracy, 0.801; recall@5, 0.731; mean reciprocal rank, 0.710) and clinician-rated reasoning scores. Higher clinician-rated diagnostic reasoning quality was strongly associated with diagnostic correctness in mixed-effects logistic regression analyses ({beta} = 1.80; p < 0.001), corresponding to an approximately six-fold increase in odds of a correct diagnosis per 1-point increase in reasoning score. In an illustrative comparison, diagnostic accuracy of Claude Opus 4.5 fell within the range observed for psychiatry trainees. Conclusions and RelevanceLLMs demonstrated high diagnostic accuracy and generated clinical reasoning that clinicians judged to be largely coherent and safe. Diagnostic reasoning quality was more strongly associated with diagnostic correctness than data extraction quality, underscoring the importance of evaluating reasoning alongside accuracy when assessing LLMs for clinical decision support in psychiatry. Key PointsO_ST_ABSQuestionC_ST_ABSCan multiple large language models accurately diagnose psychiatric conditions and generate diagnostic reasoning that clinicians judge as coherent, safe, and clinically meaningful? FindingsAcross 196 psychiatric case vignettes, four large language models demonstrated high diagnostic accuracy. In a clinician-evaluated subset of 30 vignettes, model diagnostic accuracy fell within the range observed for psychiatry residents. Clinicians judged model-generated diagnostic reasoning to be largely coherent and safe. Higher clinician-rated reasoning quality was strongly associated with diagnostic correctness, independent of data extraction quality. MeaningEvaluating diagnostic reasoning, in addition to accuracy, may be important when assessing large language models for potential clinical decision support in psychiatry.

BackgroundCerebellar transcranial magnetic stimulation (TMS) may serve as an adjuvant therapy for psychosis symptoms, most recently we have shown improvements in negative symptoms. Historically, cerebellum TMS has not utilized functional neuroanatomy for targeting, and the precision of TMS to the cerebellum is unclear. A classical view of the cerebellum as solely involved in motor computations has been updated with the discovery of rich non-motor connectivity including the default, dorsal attention, frontoparietal control and ventral attention networks. We sought to assess cerebellar TMS magnetic field effect within individually defined networks of the cerebellum. MethodsImaging data from schizophrenia and schizoaffective participants (n=27) in a double-blinded trial of cerebellar TMS (NCT05343598) was used. Individualized resting-state connectivity fMRI maps of the cerebellum was computed for 7 canonical networks (Yeo et al 2011; Buckner et al 2011). Individualized TMS simulations were computed in SimNIBS with real-world participant-specific coil placement and intensity determination. ResultsThe peak stimulation effect (99th percentile) for each network in each participant was computed. The electric field induced by cerebellar TMS predominantly engaged specific functional networks more than others (p<0.001), indicating selective targeting of these networks. The strongest effects were found on default (44.4%), limbic (37%) and frontoparietal control (11.1%) networks. Cerebellar brain network organization was found to be similar in the patient sample to previously published large-sample organization. ConclusionsFor personalized TMS, it is important to consider the targeted network, as well as the potential off-target network effects. Our findings demonstrate that cerebellar TMS has the strongest field effect on non-motor, cognitive and affective networks within the cerebellum. These results suggest cerebellar TMS may be ideal for schizophrenia symptoms unaddressed by pharmacological treatments, and effects may vary by individual network topology.