Methodological and Quality Flaws in the Use of Artificial Intelligence in Mental Health Research: Systematic Review

JMIR Mental Health, Journal Year: 2023, Volume and Issue: 10, P. e42045 - e42045

Published: Feb. 2, 2023

Artificial intelligence (AI) is giving rise to a revolution in medicine and health care. Mental conditions are highly prevalent many countries, the COVID-19 pandemic has increased risk of further erosion mental well-being population. Therefore, it relevant assess current status application AI toward research inform about trends, gaps, opportunities, challenges.This study aims perform systematic overview applications terms methodologies, data, outcomes, performance, quality.A search PubMed, Scopus, IEEE Xplore, Cochrane databases was conducted collect records use cases for disorder studies from January 2016 November 2021. Records were screened eligibility if they practical implementation clinical trials involving conditions. evaluated categorized by International Classification Diseases 11th Revision (ICD-11). Data related trial settings, collection methodology, features, model development evaluation extracted following CHARMS (Critical Appraisal Extraction Systematic Reviews Prediction Modelling Studies) guideline. Further, bias provided.A total 429 nonduplicated retrieved 129 included full assessment-18 which manually added. The distribution found unbalanced between ICD-11 categories. Predominant categories Depressive disorders (n=70) Schizophrenia or other primary psychotic (n=26). Most interventions based on randomized controlled (n=62), followed prospective cohorts (n=24) among observational studies. typically applied evaluate quality treatments (n=44) stratify patients into subgroups clusters (n=31). Models usually combination questionnaires scales symptom severity using electronic (n=49) as well medical images (n=33). Quality assessment revealed important flaws process data preprocessing pipelines. One-third (n=56) did not report any preparation. One-fifth models developed comparing several methods (n=35) without assessing their suitability advance small proportion reported external validation (n=21). Only 1 paper second previous model. Risk transparent reporting yielded low scores due poor strategy adjusting hyperparameters, coefficients, explainability models. collaboration anecdotal (n=17) mostly remained private (n=126).These significant shortcomings, alongside lack information ensure reproducibility transparency, indicative challenges that needs face before contributing solid base knowledge generation being support tool management.

Language: Английский

---

Canonical Correlation Analysis and Partial Least Squares for Identifying Brain–Behavior Associations: A Tutorial and a Comparative Study

Biological Psychiatry Cognitive Neuroscience and Neuroimaging, Journal Year: 2022, Volume and Issue: 7(11), P. 1055 - 1067

Published: Aug. 8, 2022

Canonical correlation analysis (CCA) and partial least squares (PLS) are powerful multivariate methods for capturing associations across 2 modalities of data (e.g., brain behavior). However, when the sample size is similar to or smaller than number variables in data, standard CCA PLS models may overfit, i.e., find spurious that generalize poorly new data. Dimensionality reduction regularized extensions have been proposed address this problem, yet most studies using these approaches some limitations. This work gives a theoretical practical introduction into common CCA/PLS their variants. We examine limitations variables. discuss how dimensionality regularization techniques problem explain main advantages disadvantages. highlight crucial aspects framework, including optimizing hyperparameters model testing identified statistical significance. apply described simulated real from Human Connectome Project Alzheimer's Disease Neuroimaging Initiative (both n > 500). use both low- high-dimensionality versions (i.e., ratios between range ∼1–10 ∼0.1–0.01, respectively) demonstrate impact on models. Finally, we summarize key lessons tutorial.

Language: Английский

---

Signature of Altered Retinal Microstructures and Electrophysiology in Schizophrenia Spectrum Disorders Is Associated With Disease Severity and Polygenic Risk

Biological Psychiatry, Journal Year: 2024, Volume and Issue: 96(10), P. 792 - 803

Published: April 27, 2024

Language: Английский

---

Association between Adverse Childhood Experiences and long-term outcomes in people at Clinical High-Risk for Psychosis

Schizophrenia, Journal Year: 2025, Volume and Issue: 11(1)

Published: Feb. 20, 2025

Adverse childhood experiences (ACEs) are common in people at clinical high-risk for psychosis (CHR), however, the relationship between ACEs and long-term outcomes is still unclear. This study examined associations CHR individuals. 344 individuals 67 healthy controls (HC) were assessed using Childhood Trauma Questionnaire (CTQ), Bullying Experience of Care Abuse (CECA). followed up to 5 years. Remission from state, transition (both defined with Comprehensive Assessment an At Risk Mental State), level functioning (assessed Global Functioning) assessed. Stepwise multilevel logistic regression models used investigate outcomes. significantly more prevalent than HC. Within cohort, physical abuse was associated a reduced likelihood remission (OR = 3.64, p 0.025). Separation parent linked increased both 0.32, 0.011) higher 1.77, 0.040). Death 1.87, 0.037) risk transitioning psychosis. Physical death related adverse CHR. The counter-intuitive association separation may reflect removal child environment. Future studies should whether interventions targeting effect specific might help improve this population.

Language: Английский

---

Long-term Neural Embedding of Childhood Adversity in a Population-Representative Birth Cohort Followed for 5 Decades

Biological Psychiatry, Journal Year: 2021, Volume and Issue: 90(3), P. 182 - 193

Published: March 9, 2021

Childhood adversity has been previously associated with alterations in brain structure, but heterogeneous designs, methods, and measures have contributed to mixed results impeded progress mapping the biological embedding of childhood adversity. We sought identify long-term differences structural integrity adversity.Multiple regression was used test associations between prospectively ascertained during retrospectively reported adulthood magnetic resonance imaging midlife global regional cortical thickness, surface area, subcortical gray matter volume 861 (425 female) members Dunedin Study, a longitudinal investigation population-representative birth cohort.Both were integrity, consistently stronger more widely distributed than Sensitivity analyses revealed that these not driven by any particular or category (i.e., threat deprivation) socioeconomic disadvantage. Network enrichment localized broadly along hierarchical gradient information processing.Exposure is widespread across structures, suggesting long lasting, localized. Research using likely underestimates magnitude associations. These findings may inform future research investigating mechanisms through which becomes embedded influences mental health cognition.

Language: Английский

---

Exploring causal mechanisms of psychosis risk

Neuroscience & Biobehavioral Reviews, Journal Year: 2024, Volume and Issue: 162, P. 105699 - 105699

Published: May 6, 2024

Robust epidemiological evidence of risk and protective factors for psychosis is essential to inform preventive interventions. Previous syntheses have classified these according their strength association with psychosis. In this critical review we appraise the distinct overlapping mechanisms 25 key environmental psychosis, link mechanistic pathways that may contribute neurochemical alterations hypothesised underlie psychotic symptoms. We then discuss implications our findings future research, specifically considering interactions between factors, exploring universal subgroup-specific improving understanding temporality dynamics, standardising operationalisation measurement developing interventions targeting factors.

Language: Английский

---

Altered regional brain activity and functional connectivity patterns in major depressive disorder: A function of childhood trauma or diagnosis?

Journal of Psychiatric Research, Journal Year: 2022, Volume and Issue: 147, P. 237 - 247

Published: Jan. 17, 2022

Childhood trauma (CT) is a non-specific risk factor for major depressive disorder (MDD). However, the neurobiological mechanisms of MDD with CT remain unclear. In present study, we sought to determine specific brain regions associated and etiology. Fractional amplitude low-frequency fluctuation (fALFF) functional connectivity (FC) analyses were performed assess alterations intrinsic activity in CT, without healthy controls CT. Two-by-two factorial examine effects factors "MDD" "CT" on fALFF FC. Moderator analysis was used explore whether severity depression moderated relationship between aberrant fALFF. We found that etiological exhibited negative impacts dysfunction including altered left postcentral gyrus, lingual paracentral lobule (PCL), cuneus. Decreased FC observed following regions: (i) gyrus seed fusiform as well right calcarine cortex; (ii) PCL supplementary motor area, cortex, precentral cuneus; (iii) superior parietal lobule, gyrus. Furthermore, acted moderator PCL. These data indicate patients exposure are clinically neurobiologically distinct.

Language: Английский

---

The separation distress hypothesis of depression – an update and systematic review

Neuropsychoanalysis, Journal Year: 2023, Volume and Issue: 25(2), P. 103 - 159

Published: July 3, 2023

The separation distress hypothesis of depression, formulated in 2009, offered an affective neuroscience perspective on our most common and costly mental health condition, as alternative to molecular reductionism psychiatry. Our constituted a neurobiological extension classic work depression intrinsically related attachment loss, integrating psychological findings from clinical preclinical models. We posited three ideas: 1) is evolutionarily conserved vulnerability within the social-affective endowment mammalian brains, involving interactions between other CNS systems; 2) this mechanism adaptively shuts down protracted distress, potentially lethal infant mammals, thus protective circumscribed form, explaining its conservation; 3) process has no single biological 'lever' instantiated through complex recursion. Like processes, it overlaps with kindred processes also serving behavioral shutdown, particularly hibernation sickness behavior. Epicenters for include large array systems regulating behavioral/affective states: stress axis, immune systems, monoamine cholinergic GABA glutamate molecule pro-social neuropeptide all dynamically linked. This review tracks multiple developments last 10 years including growing public healthcare awareness about not just impact primary relationships, corollary, neurobiologically destructive effects trauma, abuse, neglect, social isolation. Secure both necessity, loss or traumatic disruption may constitute fertile ground forms.

Language: Английский

---

Resting-state functional connectivity patterns associated with childhood maltreatment in a large bicentric cohort of adults with and without major depression

Psychological Medicine, Journal Year: 2022, Volume and Issue: 53(10), P. 4720 - 4731

Published: June 27, 2022

Childhood maltreatment (CM) represents a potent risk factor for major depressive disorder (MDD), including poorer treatment response. Altered resting-state connectivity in the fronto-limbic system has been reported maltreated individuals. However, previous results smaller samples differ largely regarding localization and direction of effects.We included healthy depressed [n = 624 participants with MDD; n 701 control (HC) participants] that underwent functional MRI measurements provided retrospective self-reports using Trauma Questionnaire. A-priori defined regions interest [ROI; amygdala, hippocampus, anterior cingulate cortex (ACC)] were used to calculate seed-to-voxel connectivities.No significant associations between any ROI found across MDD HC no interaction effect diagnosis became significant. Investigating patients only yielded maltreatment-associated increased amygdala dorsolateral frontal areas [pFDR < 0.001; η2partial 0.050; 95%-CI (0.023-0.085)]. This was robust various sensitivity analyses associated concurrent symptom severity. Particularly strong amygdala-frontal observed acutely individuals 264; pFDR 0.091; (0.038-0.166)). Weaker evidence - not surviving correction multiple altered supracallosal ACC maltreatment.The majority correlates CM could be replicated this large-scale study. The strongest clinically relevant adult amygdala-dorsolateral depression. Future studies should explore relevance pathway subgroup patients.

Language: Английский

---

Clinical, Brain, and Multilevel Clustering in Early Psychosis and Affective Stages

JAMA Psychiatry, Journal Year: 2022, Volume and Issue: 79(7), P. 677 - 677

Published: May 18, 2022

Approaches are needed to stratify individuals in early psychosis stages beyond positive symptom severity investigate specificity related affective and normative variation validate solutions with premorbid, longitudinal, genetic risk measures.

Language: Английский

---