Psychiatry and Clinical Neurosciences

ObjectiveTo determine the clinical relevance of brain temperature (TBr) variation in patients after traumatic brain injury (TBI). DesignCohort study with prospective (healthy participant) and retrospective (TBI patient) arms. SettingSingle neuroimaging site in the UK (prospective arm); intensive care sites contributing to the Collaborative European NeuroTrauma Effectiveness Research in TBI (CENTER-TBI) High Resolution ICU (HR ICU) Sub-Study (retrospective arm). Participants40 healthy adults aged 20-40 years recruited for non-invasive brain thermometry and all patients up to May 2020 that had TBr measured directly and were not subjected to Targeted Temperature Management (TTM). Main outcome measuresA diurnal change in TBr (healthy participants); death in intensive care (patients). ResultsIn healthy participants, mean TBr (38.5 SD 0.4{degrees}C) was higher than oral temperature (36.0 SD 0.5{degrees}C), and 0.36{degrees}C higher in luteal females relative to follicular females and males (95% confidence interval 0.17 to 0.55, P=0.0006 and 0.23 to 0.49, P<0.0001, respectively). TBr increased with age, most notably in deep brain regions (0.6{degrees}C over 20 years; 0.11 to 1.07, P=0.0002). The mean maximal spatial TBr range was 2.41 (SD 0.46){degrees}C, with highest temperatures in the thalamus. TBr varied significantly by time of day, especially in deep brain regions (0.86{degrees}C; 0.37 to 1.26, P=0.0001), and was lowest in the late evening. Diurnal TBr in cortical white matter across participants ranged from 37.0 to 40.3{degrees}C. In TBI patients (n=114), mean TBr (38.5 SD 0.8{degrees}C) was significantly higher than body temperature (TBo 37.5 SD 0.5{degrees}C; P<0.0001) and ranged from 32.6 to 42.3{degrees}C. Only 25/110 patients displayed a diurnal temperature rhythm; TBr amplitude was reduced in older patients (P=0.018), and 25/113 patients died in intensive care. Lack of a daily TBr rhythm, or an age increase of 10 years, increased the odds of death 12-fold and 11-fold, respectively (OR for death with rhythm 0.09; 0.01 to 0.84, P=0.035 and for death with ageing by 1 year 1.10; 1.05 to 1.16, P=0.0002). Mean TBr was positively associated with survival (OR for death 0.45 for 1{degrees}C increase; 0.21 to 0.96, P=0.040). ConclusionsHealthy TBr exceeds TBo and varies by sex, age, menstrual cycle, brain region, and time of day. Our 4-dimensional reference resource for healthy TBr can guide interpretation of TBr data in multiple clinical settings. Daily temperature variation is frequently disrupted or absent in TBI patients, in which TBr variation is of greater prognostic use than absolute TBr. Older TBI patients lacking a daily TBr rhythm are at greatest risk of death in intensive care. Appropriately controlled trials are needed to confirm the predictive power of TBr rhythmicity in relation to patient outcome, as well as the clinical utility of TTM protocols in brain-injured patients. RegistrationUK CRN NIHR CPMS 42644; ClinicalTrials.gov number, NCT02210221. SUMMARY BOXO_ST_ABSWhat is already known on this topicC_ST_ABSO_LIBrain temperature (TBr) can be measured directly in brain-injured patients via intracranial probe, but this method cannot be used in healthy individuals. C_LIO_LITBr can be measured non-invasively using magnetic resonance spectroscopy (MRS), but this method is not appropriate for most brain-injured patients. C_LIO_LISince physiological reference ranges for TBr in health have not been established, the clinical relevance of TBr variation in patients is unknown, and the use of TTM in neurocritical care remains controversial. C_LI What this study addsO_LIA reference map for healthy adult TBr at three clinically-relevant time points that can guide interpretation of TBr measured directly, or by MRS, in multiple clinical settings. C_LIO_LIOur results suggest that loss of diurnal TBr rhythmicity after TBI increases the odds of intensive care death 12-fold; some TTM strategies may be clinically inappropriate. C_LI

The current clinical diagnosis of psychiatric disorders relies heavily on subjective assessment of symptoms. While neuroimaging has made an essential contribution to characterizing the brain of psychiatric disorders, it does not currently serve the clinical diagnosis of major psychiatric disorders. Here, we report a neuroimaging-aided diagnostic system for major psychiatric disorders designed for clinical needs. We developed novel deep learning networks with attentional mechanisms and applied them to a large-scale, single-center neuroimaging dataset containing four major psychiatric disorders and healthy groups (n=2490). Both cross-validation and extensive independent validation using multiple open-source datasets (n = 1972) showed that the system could accurately identify any one of the four diagnostic categories and healthy population from brain structural imaging. For the first time, we have constructed an automatic neuroimaging-aid diagnostic system that considers common issues in practice, such as co-morbid diagnoses and the discrimination between specific suspected diagnoses. Furthermore, real-world applications have validated the systems effectiveness. These works contribute to the translation of brain research to objective diagnostic aids for psychiatric disorders.

ObjectiveTo examine the causal relationship between alcohol use and dementia risk across multiple ancestry groups. DesignWe triangulated evidence from observational and univariable and multivariable Mendelian randomization. Setting and participantsCross-ancestry observational analyses were conducted in two large prospective studies: the US Million Veteran Program and UK Biobank. One- and two-sample univariable and multivariable Mendelian randomization used de novo data from a genome-wide association study in Million Veteran Program plus publicly available data. Main outcome measureAll-cause dementia. ResultsAmong 559,559 participants (aged 56-72 years old at baseline) included in observational analyses, 14,540 received developed dementia and 48,034 died during follow-up. Observational associations between alcohol and dementia were U-shaped. Non-, heavy (>40 drinks per week - hazard ratio (HR)1.41; 95% confidence interval [CI], 1.15 to 1.74]) and dependent (1.51[1.42-1.60]) drinkers were at higher dementia risk than light drinkers. In contrast, genetic analyses revealed a monotonically increasing association between alcohol dose and dementia, with no evidence supporting a protective effect of any level of drinking. A two-fold increase in genetically-predicted alcohol use disorder prevalence was associated with a 16% increase in dementia cases (IVW OR=1.16[1.03-1.30]), and a one standard deviation increase in log-transformed drinks per week was associated with a 15% increase (IVW OR=1.15[1.03-1.27]). ConclusionsAlcohol consumption has a causal role for dementia. These findings challenge a purported protective effect of moderate drinking. Reducing alcohol use could be an effective dementia prevention strategy.

This study anonymously examined 2,734 psychiatric patients worldwide for worsening of their pre-existing psychiatric condition during the COVID-19 pandemic. Valid responses mainly from 12 featured countries indicated self-reported worsening of psychiatric conditions in 2/3rd of the patients assessed that was validated through their significantly higher scores on scales for general psychological disturbance, post-traumatic stress disorder, and depression. Female gender, feeling no control of the situation and reporting dissatisfaction with the response of the state during the COVID-19 pandemic, and reduced interaction with family and friends increased the worsening of pre-existing psychiatric conditions, whereas optimism, ability to share concerns with family and friends and using social media like usual were associated with less worsening. An independent clinical investigation from the USA confirmed worsening of psychiatric conditions during the COVID-19 pandemic based on identification of new symptoms that necessiated clinical interventions such as dose adjustment or starting new medications in more than half of the patients.

Delirium is a neurologic syndrome characterized by inattention and cognitive impairment frequently encountered in the medically ill. Peripheral inflammation is a key trigger of delirium, but the patient-specific immune responses associated with delirium development and resolution are unknown. This retrospective cohort study of prospectively collected biospecimens examines RNA sequencing from peripheral blood mononuclear cells of adults hospitalized for COVID-19 to better understand patient-specific factors associated with delirium (n = 64). Longitudinal transcriptomic analyses highlight persistent immune dysregulation in delirium, marked by increasing expression trajectories of genes linked to innate immune pathways, including complement activation, cytokine production, and monocyte/macrophage recruitment. Genes involved adaptive immunity showed a declining trajectory over time in patients with delirium. Although corticosteroid treatment suppressed some aspects of immune hyperactivation, aberrant responses contributing to delirium were exacerbated. Delirium resolution was characterized by normalization of key transcripts such as CCL2 and innate immune markers. Novel associations with delirium were found in genes related to stress granule assembly and DUSP2 and KLF10, which mediate T-cell responses. These findings provide insights into the peripheral immune responses accompanying delirium and their modulation by corticosteroids. Future trials targeting aberrant inflammatory responses may mitigate the severe outcomes associated with delirium due to COVID19.

BackgroundWe investigated delirium prevalence and potential effects of long-term sedation in critically ill COVID-19 patients; to identify opportunities for improving sedation practices and delirium prevention. MethodsThis prospective, single-center, observational cohort study was conducted from April-June 2020. Adult COVID-19 patients were eligible if admitted to an ICU with mechanical ventilation/intravenous sedation; or a general care unit with brain monitoring due to altered mental status. Patients were evaluated daily until discharge using the Richmond Agitation-Sedation Scale, Confusion Assessment Method for the ICU, and CAM-Severity. Cumulative doses of sedation and paralytic medications were recorded. At three months post-enrollment, cognition, mood, and quality of life were measured by the Telephone Interview for Cognitive Status (TICS), Center for Epidemiologic Studies Depression Scale 10-item (CES-10), and EuroQol 5-Dimension-3 Level (EQ-5D-3L), respectively. Results67 patients were enrolled, with a mean (SD) age of 59 (12) years, 30 (45%) Hispanic, 43 (64%) developing acute respiratory distress syndrome, 55 (82%) mechanically ventilated (mean duration of 22.9 days), and 5 comatose for the entire study. Of the 62 patients assessed for delirium, 61 (98%) had delirium at least once, with a mean (SD) of 12.7 (13.0) days. >90% of patients received opioids, benzodiazepines, or propofol at least once; median (IQR) total dose of 37.4 (78.9) mg (fentanyl equivalents), 52.5 (813.3) mg (midazolam equivalents), and 46 (53) g (propofol), respectively. At follow-up, 40 (60%) patients were reached, while 16 (24%) were deceased/comfort measures. Patients showed reductions in cognition, mood, and quality of life with median (IQR) scores for TICS (0-41): 30 (26-33); CES-D-10 (0-30): 6 (4-12); EQ-5D-3L (1-3): 2 (mobility, self-care, usual activities, pain/discomfort). ConclusionCritically and acutely ill patients with COVID-19 early in the pandemic experienced a high rate of delirium and sedation. Large doses of sedatives may contribute to greater delirium burden during hospitalization, and lead to poor clinical outcomes.

Despite a global decrease in suicide rates in recent years, death by suicide has increased in the United States. It is therefore imperative to identify the risk factors associated with suicide attempts in order to combat this growing epidemic. In this study, we use an explainable-artificial intelligence method, iterative Random Forest, to predict suicide attempts using data from the Million Veteran Program. Our predictive model incorporates multiple environmental variables (e.g., elevation, light wavelength absorbance, temperature, humidity, etc) at ZIP code-level geospatial resolution. We additionally consider demographic variables from the American Community Survey as well as the number of firearms and alcohol vendors per 10,000 people in order to assess the contributions of proximal environment, access to means, and restraint decrease to suicide attempts. Our results show that geographic areas with higher concentrations of married males living with spouses are predictive of lower rates of suicide attempts, whereas geographic areas where males are more likely to live alone and to rent housing are predictive of higher rates of suicide attempts. We also identified climatic features that were associated with suicide attempt risk by age group. Additionally, we observed that firearms and alcohol vendors were associated with increased risk for suicide attempts irrespective of the age group examined, but that their effects were small in comparison to the top features. Taken together, our findings highlight the importance of social determinants and environmental factors in understanding suicide risk among veterans.

ObjectiveDetermine whether acute behavioral, electroencephalography (EEG), and functional MRI (fMRI) biomarkers of consciousness are associated with outcome after severe traumatic brain injury (TBI). MethodsPatients with acute severe TBI admitted consecutively to the intensive care unit (ICU) participated in a multimodal battery assessing behavioral level of consciousness (Coma Recovery Scale-Revised [CRS-R]), cognitive motor dissociation (CMD; task-based EEG and fMRI), covert cortical processing (CCP; stimulus-based EEG and fMRI), and default mode network connectivity (DMN; resting-state fMRI). The primary outcome was 6-month Disability Rating Scale (DRS) total scores. ResultsWe enrolled 55 patients with acute severe TBI. Six-month outcome was available in 45 (45.2{+/-}20.7 years old, 70% male), of whom 10 died, all due to withdrawal of life-sustaining treatment (WLST). Behavioral level of consciousness and presence of command-following in the ICU were each associated with lower (i.e., better) DRS scores (p=0.003, p=0.011). EEG and fMRI biomarkers did not strengthen this relationship, but higher DMN connectivity was associated with better recovery on multiple secondary outcome measures. In a subsample of participants without command-following on the CRS-R, CMD (EEG:18%; fMRI:33%) and CCP (EEG:91%; fMRI:79%) were not associated with outcome, an unexpected result that may reflect the high rate of WLST. However, higher DMN connectivity was associated with lower DRS scores ({rho}[95%CI]=-0.41[-0.707, -0.027]; p=0.046) in this group. InterpretationStandardized behavioral assessment in the ICU may improve prediction of recovery from severe TBI. Further research is required to determine whether integrating behavioral, EEG, and fMRI biomarkers of consciousness is more predictive than behavioral assessment alone.

IntroductionMeditation is widely used to support mental well-being, and recent randomized trials suggest benefits for persistent psychotic symptoms in schizophrenia. However, the magnitude and timing of causal treatment effects, response heterogeneity, and underlying neurobiological mechanisms over clinically meaningful timescales remain unclear. MethodsWe analyzed data from an eight-month, parallel-group randomized clinical trial (ChiCTR1800014913) of 64 male inpatients with chronic schizophrenia, randomized to daily clinician-guided meditation plus rehabilitation or rehabilitation alone. Prespecified outcomes were PANSS percentage decrease rate and RBANS increase rate. Linear mixed-effects models estimated time-specific causal average treatment effects. Cross-lagged panel models examined temporal relations between symptom and cognitive benefits; latent-class mixed models characterized treatment-response heterogeneity. Resting-state fMRI at baseline, 3, and 8 months yielded functional components, their complexity indices, and functional-connectivity predictors of clinical benefit. ResultsMeditation produced progressive symptom improvement (average treatment effects on PANSS decrease rate: 11.8% after 3 months; 20.8% after 8 months) and an early cognitive gain (7.6% after 3 months) that plateaued. Early cognitive improvement predicted, but did not mediate, later symptom relief. Response trajectories were heterogeneous; marital status and lower antipsychotic burden characterized high responders. Neuroimaging revealed a biphasic pattern: higher baseline default-mode-cerebellar complexity predicted short-term benefit, whereas greater 3-month action-mode-sensorimotor-executive complexity predicted longer-term gains; functional-connectivity models converged on these findings. ConclusionsClinician-guided meditation, added to rehabilitation, yields robust causal treatment effects on symptoms in schizophrenia. A biphasic shift from default-mode-cerebellar involvement to action-mode engagement provides phase-specific, information-based indicators to guide personalized meditation in severe mental illness.

Schizophrenia spectrum disorders (SSDs) are clinically and neurobiologically heterogeneous. Normative modeling addresses heterogeneity of structural brain alterations by focusing on individual-level deviations, but their clinical relevance in SSDs remains controversial. We mapped the relationship between individual gray matter volume (GMV) deviations and schizophrenia diagnosis and symptoms. Normative models of GMV were established using cross-sectional, T1-weighted magnetic resonance imaging data from a large, multi-site, healthy reference cohort (N = 7957). Deviations were derived for SSD patients (n = 379) and healthy controls (n =149). Patients showed a significantly more negative average deviation compared to controls and regional deviations predicted diagnostic status with adequate performance (AUC = 0.79). A more negative deviation was associated with higher symptom severity and lower cognitive functioning in SSD. Negative deviations were scattered across the brain, with the largest alterations in the salience network. Our findings strengthen the potential of normative modeling to disentangle the heterogeneous underpinnings of SSD and provide further evidence for individualized structural deviations, particularly in the salience network, as promising markers of illness severity in SSDs.

BackgroundIn the last decades, advances in Intensive Care Unit management have led to decreased mortality. However, significant morbidity remains as patients survive after a lesional coma with uncertain quality of awakening and high risk of functional disability. Predicting this level of recovery but also the functional disability of those who will awake constitutes a major challenge for medical, ethical and social perspectives. Among the huge heterogeneity of coma-related injuries, recognising the universality of a common functional pattern which may be focused on a final step of an integrated network would be of great interest for our understanding of disorders of consciousness. The objective of this study is to investigate the neural correlates of arousal and awareness in coma and post-coma to build a prognostic tool based on the detection of a common pattern between patients with a favourable versus an unfavourable outcome. Method/DesignWe will implement this objective in a translational approach which combines PET-MR imaging, neurophysiology, behavioural/clinical assessments and innovative statistical and computational analysis tools in patients with disorders of consciousness in Intensive Care Unit and in Rehabilitation Department.

BackgroundPost-critical illness cognitive dysfunction (PCICD) is a common and debilitating condition affecting survivors of critical illness. While sepsis has been implicated in poor cognitive outcomes, its independent contribution remains unclear due to multiple associated confounders in critical illness. This study aimed to characterize cognitive recovery trajectories over 12 months post-intensive care unit (ICU) and to evaluate the influence of sepsis and benzodiazepine exposure on cognitive outcomes. MethodsIn this single-center, prospective cohort study, adult ICU survivors were assessed at 30 days, 3 months, 6 months, and 12 months post-discharge using the telephone-administered Mini-Mental State Examination (MMSE) or Montreal Cognitive Assessment (MoCA). Scores were standardized into z-scores for comparability. Mixed-effects models assessed changes over time and the effects of clinical covariates, including sepsis status and benzodiazepine exposure. Additionally, we investigated whether any one specific cognitive domain was disproportionally impaired by critical illness over time. ResultsOf 197 eligible patients during the enrollment period, 77 (39%) completed at least one cognitive assessment. Standardized cognitive scores significantly improved over time, with the greatest gains observed within 6 months: +0.40 SD at 3 months (p = 0.041), +0.54 SD at 6 months (p = 0.016), and +0.49 SD at 12 months (p = 0.033) compared to scores at the time of acute illness. Sepsis status had no significant effect on recovery trajectory. No single cognitive domain was disproportionately affected by critical illness; instead, changes were observed in the overall score over time. Benzodiazepine exposure showed complex associations: longer duration (-0.24 SD/day, p = 0.008) and higher daily dose (-0.02 SD/unit, p = 0.006) were linked to worse cognition, while total cumulative dose was paradoxically associated with better scores (+0.03 SD/unit, p < 0.001), possibly reflecting confounding by indication or survival bias. ConclusionsICU survivors experience gradual cognitive recovery over the first year, primarily within 6 months. Sepsis does not independently affect this trajectory. Benzodiazepine exposure, especially prolonged or high daily dosing, emerges as a modifiable risk factor for cognitive impairment, consistent with prior investigations of PCICD. These findings highlight the importance of sedation strategies and structured cognitive follow-up.

A primary aim of precision psychiatry is the establishment of predictive models linking individual differences in brain functioning with clinical symptoms. In particular, cognitive impairments are transdiagnostic, treatment resistant, and contribute to poor clinical outcomes. Recent work suggests thousands of participants may be necessary for the accurate and reliable prediction of cognition, calling into question the utility of most patient collection efforts. Here, using a transfer-learning framework, we train a model on functional imaging data from the UK Biobank (n=36,848) to predict cognitive functioning in three transdiagnostic patient samples (n=101-224). The model generalizes across datasets, and brain features driving predictions are consistent between populations, with decreased functional connectivity within transmodal cortex and increased connectivity between unimodal and transmodal regions reflecting a transdiagnostic predictor of cognition. This work establishes that predictive models derived in large population-level datasets can be exploited to boost the prediction of cognitive function across clinical collection efforts.

Traumatic brain injury (TBI) is a leading cause of disability worldwide. Yet, our understanding of the mechanisms of this condition is limited, especially in the acute setting. Here, we investigated the relationship between functional connectivity and common clinical assessments, like the Glasgow Coma Scale (GCS) at admission and modified Rankin scale at 3-months (mRS) to determine if functional connectivity can provide a broader representation of the brains networks than these standard tests. We performed a retrospective analysis of resting state functional MRI and clinical data in 58 patients (41.28 {+/-} 18.63) scanned acutely/subacutely ([&le;] 31 days). Then, for a secondary analysis, we included 50 more patients who presented after either a first or a repeat incident and were scanned either acutely/subacutely or chronically (<2 yrs) (all together 108 patients, 46.4 {+/-} 20.1yrs). Using a 268-node functional atlas, we derived 35,778 unique edges, based on which we calculated the mean functional connectivity of 10 resting state networks and used those to establish a link to TBI severity and functional outcome. Our analysis showed that when dividing sub/acute patients (n=58) based on GCS severity, only the Subcortical network showed a significant discrimination between mild and moderate-severe GCS at admission (P<0.001), with hyperconnectivity noted in mild patients, and hypoconnectivity - in moderate-severe GCS patients. This difference appeared to be mainly driven by the thalami (Right, P=0.002; Left P<0.001). Similar results were observed when investigating GCS subscores at admission (Eyes, Motor, Verbal, all P<0.001). Further, when evaluating mRS outcomes at 3-months against functional connectivity, differences were noted within the Motor, Cerebellum and Medial-Frontal networks, though none survived multiple comparisons. Importantly, we found the DMN and mRS to be correlated but with a limited relationship (r2= 0.18). Lastly, we performed a post-hoc analysis (n=108) to investigate if the hyperconnectivity in the Subcortical network of sub/acute mild GCS patients remained irrespective of acuity of scanning (i.e. acute/ chronic) or frequency of TBI (i.e. first/ repeat). Our analysis showed that GCS severity appeared to be the main driver of functional connectivity within the Subcortical network, whereas acuity of scanning, alongside GCS severity contributed to the results of chronically scanned patients. While GCS and 3-month mRS scores offer some meaningful insights, their limited capture of the neural representation underscores the need to investigate whether other early clinical assessments correlate more robustly with early resting state networks or whether such networks themselves could predict future outcomes.

Background and HypothesisIndividuals with schizophrenia spectrum disorders (SSD) often show diminished reward pursuit, whereas loss avoidance is relatively preserved. The neural mechanisms of this dissociation and its relation to negative symptoms remain unclear. We hypothesized that in SSD, cognitive resources are preferentially directed toward avoiding losses rather than pursuing rewards, potentially limiting reward processing and contributing to negative symptoms. Study DesignUsing computational modeling of behavior during a two-stage decision task which distinguished between goal-directed (model-based) and habitual (model-free) strategies under reward and loss conditions, we studied 42 stable individuals with SSD and 48 healthy controls (HC) during functional magnetic resonance imaging. Study ResultsIn individuals with SSD, model-based control was shifted toward loss avoidance relative to HC, with corresponding changes in prefrontal circuitry. In anterior cingulate, orbitofrontal, and dorsolateral prefrontal regions, individuals with SSD showed increased activation during model-based control in the loss condition. Within this group, loss-biased activation in the right anterior cingulate region was associated with anhedonia. In 25 patients with available follow-up data, loss-biased activation in the right anterior cingulate region at baseline was prospectively related to worsening of motivation and social engagement over the subsequent year. ConclusionsOur findings suggest that, compared to HC, those with SSD allocate their limited cognitive resources more toward loss avoidance relative to reward pursuit. The association between loss-biased anterior cingulate engagement and anhedonia supports a neurocomputational account of diminished pleasure in psychotic disorders, with potential implications for developing motivation-targeted treatments and early prediction of negative-symptom worsening.

BackgroundStudies applying machine learning to obsessive-compulsive disorder (OCD) typically report accuracy in homogeneous samples but rarely assess model reliability, generalizability, and interpretability needed for clinical use. MethodsWe applied a transformer-based deep learning model, the Multi-Band Brain Net, to the ENIGMA-OCD cohort - the largest available resting-state functional magnetic resonance imaging (rs-fMRI) dataset in OCD with 1,706 participants (869 cases with OCD, 837 controls) across 23 sites worldwide. We evaluated model reliability by calculating calibration - the models ability to "know what it doesnt know". We assessed generalizability using leave-one-site-out validation to test performance on unseen sites with different scanners, acquisition protocols, and patient populations. Finally, we examined interpretability by analyzing model attention weights to identify the neural connectivity patterns that influence model predictions. ResultsThe model achieved modest but competitive classification performance (AUROC = .653 {+/-} .039). Crucially, while large-scale pretraining on the UK Biobank (N = 40,783) did not boost accuracy, it significantly enhanced model calibration by reducing overconfident predictions. Leave-one-site-out validation showed a generalization gap across sites (AUROC = .427-.819). Pretraining did not close this gap but removed scanner manufacturer bias. Finally, attention-based mapping identified biologically plausible patterns of widespread hypoconnectivity in OCD relative to healthy controls, particularly in low-frequency bands involving the default mode, salience, and somatomotor networks. These findings aligned with known OCD neurobiology. ConclusionsThis study provides a framework for developing more reliable and trustworthy clinical artificial intelligence for OCD.

Molecular neuroimaging techniques, like PET and SPECT, offer invaluable insights into the brains in-vivo biology and its dysfunction in neuropsychiatric patients. However, the transition of molecular neuroimaging into diagnostics and precision medicine has been limited to a few clinical applications, hindered by issues like practical feasibility and high costs. In this study, we explore the use of normative modelling (NM) for molecular neuroimaging to identify individual patient deviations from a reference cohort of subjects. NM potentially addresses challenges such as small sample sizes and diverse acquisition protocols that are typical of molecular neuroimaging studies. We applied NM to two PET radiotracers targeting the dopaminergic system ([11C]-(+)-PHNO and [18F]FDOPA) to create a normative model to reference groups of controls. The models were subsequently utilized on various independent cohorts of patients experiencing psychosis. These cohorts were characterized by differing disease stages, treatment responses, and the presence or absence of matched controls. Our results showed that patients exhibited a higher degree of extreme deviations ([~]3-fold increase) than controls, although this pattern was heterogeneous, with minimal overlap in extreme deviations topology (max 20%). We also confirmed the value of striatal [18F]FDOPA signal to predict treatment response (striatal AUC ROC: 0.77-0.83). Methodologically, we highlighted the importance of data harmonization before data aggregation. In conclusion, normative modelling can be effectively applied to molecular neuroimaging after proper harmonization, enabling insights into disease mechanisms and advancing precision medicine. The method is valuable in understanding the heterogeneity of patient populations and can contribute to maximising cost efficiency in studies aimed at comparing cases and controls.

BackgroundBrain disorders are leading contributors to increasing disability and spending worldwide. In 2022 the Swiss Brain Health Plan (SBHP) was launched to promote brain health and prevent brain disorders. To guide the implementation of the SBHP, we performed a detailed analysis of the health and economic burden of brain disorders in Switzerland. MethodsWe analyzed Global Burden of Disease 2023 disability-adjusted life years (DALYs) and Institute for Health Metrics and Evaluation (IHME) cause-specific health-care spending estimates for Switzerland. DALYs were quantified for 1990 - 2023. Spending was analyzed for 2000 - 2019 across six types of care. We examined age and sex patterns, spending distribution, and international comparisons with six other countries (Germany, France, Denmark, Norway, Italy, Singapore). To assess short- and longer-term association between burden and spending estimates, we fitted panel regression models with disorder and year fixed effects under one-year and five-year lag specifications. FindingsBoth disease burden and spending were highly concentrated in a small number of conditions in Switzerland. In 2023, ten brain disorders accounted for 82{middle dot}9% of Switzerlands total DALY burden. In 2019, ten brain disorders accounted for 86{middle dot}0% of all direct brain-health spending, with dementia alone comprising 29{middle dot}5% of total expenditures. Among seven analyzed comparator countries, Switzerland had the highest per-capita brain-health spending and the highest spending per DALY. In fixed-effects panel models that accounted for spending persistence, lagged DALYs were not statistically associated with subsequent spending. Suicide prevention and addiction showed significant lower-than-expected health-sector spending (self-harm: {beta} = -0{middle dot}23; drug use disorders: {beta} = -0{middle dot}08 to -0{middle dot}18 across lag models). InterpretationBrain disorders generate a large burden in Switzerland. Within the IHME estimates, the burden-spending relationship over time appears limited. The implementation of the SBHP will refer to the current data and call for a burden-informed financing to guide strategic cross-sectorial allocation and prevention investments. Research in contextO_ST_ABSEvidence before this studyC_ST_ABSWe drew on evidence from the Global Burden of Disease (GBD) 2023 estimates on neurological and mental health conditions, and on cause-specific health-care spending data from the Institute for Health Metrics and Evaluation for Switzerland and selected high-income countries. These sources show that brain disorders are major contributors to disability and premature mortality, and that Switzerland is among the worlds highest spenders per capita on health care. Prior work has described the costs of individual brain disorders and drivers of health expenditure growth; however, it has often treated burden and spending as partly separate domains, leaving the country-level link between cause-specific disability-adjusted life-years (DALYs) and cause-specific spending, over time and in either direction, poorly characterized. Added value of this studyTo our knowledge, this is the first study in a single country to systematically link cause-specific DALYs and cause-specific direct health-care spending for brain disorders and to examine their longitudinal and bi-directional associations. Using harmonized GBD 2023 estimates and IHME 2019 cause-specific health spending data, we quantify the health and economic burden of 23 brain disorders in Switzerland across age, sex, care setting, and time, and benchmark patterns against six other high-income countries. By applying panel regression models with disorder and year fixed effects, we assess whether modeled spending shows any association with prior modeled burden once spending persistence is accounted for, and identify conditions with higher or lower spending relative to burden. Implications of all the available evidenceSwitzerland bears a major burden of brain disorders and devotes substantial resources to their care, yet within the modeled estimates, spending does not consistently correspond to burden over time. Disorders with long-standing multisectoral programs tended to show lower spending-to-burden ratios, suggesting that coordinated action beyond the health sector may reduce downstream health-sector demand. For Switzerland and similar health systems, these findings support national brain-health strategies that strengthen life-course prevention and early intervention, and that integrate financing with burden data to inform priority setting and periodic reassessment of resource allocation.

Background: Concerns about "AI psychosis" have swirled in the media since ChatGPT's release, but few systematic analyses exist. We therefore conducted an electronic health record (EHR) analysis to identify the frequency, clinical characteristics, and quality of AI interactions in patients experiencing psychosis treated in a medical center. Methods: AI keywords (e.g., ChatGPT, AI) were used to search Vanderbilt University Medical Center's EHR from 12/1/2022-4/1/2026. Records were discarded if they were not AI-related or if the primary diagnosis did not include psychosis. Three raters read notes to determine if a patient was experiencing AI psychosis and classified the interactions using 4 a-priori categories (Catalyst, Amplifier, Co-Author, Object) formulated to explain how AI-related negative outcomes emerge. Findings: 73 patients met our criteria. 28 patients were rated as experiencing AI psychosis, 17 had neutral interactions, and 28 expressed delusional content related to AI without documented evidence of conversational AI use. ChatGPT was the matching keyword for 53.6% patients experiencing AI psychosis. The majority of AI psychosis cases were documented after ChatGPT's "4o" model was released in May 2024. Notably, the AI Psychosis group had significantly more patients experiencing a first psychotic episode (60.7%) compared to the other two groups. Amplifier was the most common (64.3%) qualitative rating in the AI Psychosis group. Interpretation: "AI psychosis" is an infrequent but real phenomenon observed in clinical practice. Most affected patients were experiencing their first psychotic episode and presented with AI psychosis following the release of the more sycophantic GPT-4o. Among the affected patients, AI most often exacerbated an existing condition by reinforcing distorted ideas.

ImportanceWhile there is a general consensus that functional connectome pathology is a key mechanism underlying psychosis spectrum disorders, the literature is plagued with inconsistencies and translation into clinical practice is non-existent. This is perhaps because group-level findings may not be accurate reflections of pathology at the individual patient level. ObjectiveTo characterize inter-individual heterogeneity in functional networks and investigate if normative values can be leveraged to identify biologically less heterogeneous subgroups of patients. Design, Setting, and ParticipantsWe used data collected in a case-control study conducted at the University of Alabama at Birmingham (UAB). We recruited antipsychotic medication-naive first-episode psychosis patients from UAB outpatient, inpatient, and emergency room settings. Main Outcome(s) and Measure(s)Individual-level patterns of deviations from a normative reference range in resting-state functional networks using the Yeo-17 atlas for parcellations. ResultsStatistical analyses included 108 medication-naive first-episode psychosis patients. We found that there is a high level of inter-individual heterogeneity in resting-state network connectivity deviations from the normative reference range. Interestingly 48% of patients did not have any functional connectivity deviations, and no more than 11.1% of patients shared functional deviations between the same regions of interest. In a post hoc analysis, we grouped patients based on deviations into four theoretically possible groups. We discovered that all four groups do exist in our experimental data and showed that subgroups based on deviation profiles were significantly less heterogeneous compared to the overall group (positive deviation group: z= -2.88, p = 0.002; negative deviation group: z= -3.36, p<0.001). Conclusions and RelevanceOur findings experimentally demonstrate that there is a high level of inter-individual heterogeneity in resting-state network pathology in first-episode psychosis patients which support the idea that group-level findings are not accurate reflections of pathology at the individual level. We also demonstrated that normative functional connectivity deviations may have utility for identifying biologically less heterogeneous subgroups of patients, even though they are not distinguishable clinically. Our findings constitute a significant step towards making precision psychiatry a reality, where patients are selected for treatments based on their individual biological characteristics. KEY POINTSO_ST_ABSQuestionC_ST_ABSHow heterogeneous is individual-level resting-state functional network pathology in patients suffering from a first psychotic episode? Can normative reference values in functional network connectivity be leveraged to identify biologically more homogenous subgroups of patients? FindingsWe report that functional network pathology is highly heterogeneous, with no more than 11% of patients sharing functional deviations between the same regions of interest. MeaningNormative modeling is a tool that can map individual neurobiological differences and enables the classification of a clinically heterogenous patient group into subgroups that are neurobiologically less heterogenous.