Predicting individual cases of major adolescent psychiatric conditions with artificial intelligence

Translational Psychiatry, Journal Year: 2023, Volume and Issue: 13(1)

Published: Oct. 10, 2023

Abstract Three-quarters of lifetime mental illness occurs by the age 24, but relatively little is known about how to robustly identify youth at risk target intervention efforts improve outcomes. Barriers knowledge have included obtaining robust predictions while simultaneously analyzing large numbers different types candidate predictors. In a new, large, transdiagnostic sample and multidomain high-dimension data, we used 160 predictors encompassing neural, prenatal, developmental, physiologic, sociocultural, environmental, emotional cognitive features leveraged three machine learning algorithms optimized with novel artificial intelligence meta-learning technique predict individual cases anxiety, depression, attention deficit, disruptive behaviors post-traumatic stress. Our models tested well in unseen, held-out data (AUC ≥ 0.94). By utilizing large-scale design advanced computational approaches, were able compare relative predictive ability neural versus psychosocial principled manner found that consistently outperformed metrics their deliver cases. We deep networks tree-based XGBoost logistic regression ElasticNet, supporting conceptualization illnesses as multifactorial disease processes non-linear relationships among can be modeled psychiatry techniques. To our knowledge, this first study test these gold-standard from classes across multiple health conditions within same >100 Further research suggested explore findings longitudinal validate results an external dataset.

Language: Английский

---

The computational psychiatry of antisocial behaviour and psychopathy

Neuroscience & Biobehavioral Reviews, Journal Year: 2022, Volume and Issue: 145, P. 104995 - 104995

Published: Dec. 16, 2022

Antisocial behaviours such as disobedience, lying, stealing, destruction of property, and aggression towards others are common to multiple disorders childhood adulthood, including conduct disorder, oppositional defiant psychopathy, antisocial personality disorder. These have a significant negative impact for individuals society, but whether they represent clinically different phenomena, or simply approaches diagnosing the same underlying psychopathology is highly debated. Computational psychiatry, with its dual focus on identifying classes disorder health (data-driven) latent cognitive neurobiological mechanisms (theory-driven), well placed address these questions. The elucidation that might characterise processes across behaviour can also provide important advances. In this review, we critically discuss contribution computational research our understanding various disorders, highlight suggestions how psychiatry clinical scientific questions about in future.

Language: Английский

---

Machine learning reveals sex differences in distinguishing between conduct-disordered and neurotypical youth based on emotion processing dysfunction

BMC Psychiatry, Journal Year: 2025, Volume and Issue: 25(1)

Published: Feb. 6, 2025

Abstract Background Theoretical models of conduct disorder (CD) highlight that deficits in emotion recognition, learning, and regulation play a pivotal role CD etiology. With being more prevalent boys than girls, various theories aim to explain this sex difference. The “differential threshold” hypothesis suggests greater dysfunction conduct-disordered girls boys, but previous research using conventional statistical analyses has failed support hypothesis. Here, we used novel analytic techniques such as machine learning (ML) uncover potentially sex-specific differences among with compared their neurotypical peers. Methods Multi-site data from 542 youth 710 controls (64% 9–18 years) who completed tasks were analyzed multivariate ML classifier distinguish between separately by sex. Results Both female male classifiers accurately predicted (above chance level) individual status based solely on the neurocognitive features dysfunction. Notably, outperformed identifying individuals CD. However, classification identification performance both was below clinically relevant 80% accuracy threshold (although they still provided relatively fair realistic estimates ~ 60% performance), probably due substantial heterogeneity within large diverse, multi-site sample (and controls). Conclusions These findings confirm close association sexes, stronger observed affected which aligns also underscore CD, namely only subset those are likely have other domains (not tested here) contribute Clinical trial number Not applicable.

Language: Английский

---

Predicting the onset of mental health problems in adolescents

Psychological Medicine, Journal Year: 2025, Volume and Issue: 55

Published: Jan. 1, 2025

Abstract Objective Mental health problems are the major cause of disability among adolescents. Personalized prevention may help to mitigate development mental problems, but no tools available identify individuals at risk before they require care. Methods We identified children without baseline with six different clinically relevant 1- or 2-year follow-up in Adolescent Brain Cognitive Development (ABCD) study. used machine learning analysis predict these use demographic, symptom and neuroimaging data a discovery (N = 3236) validation 3851) sample. The sample 168–513 per group) consisted participants MRI were matched healthy controls on age, sex, IQ, parental education level. 84–231) data. Results Subclinical symptoms 9–10 years age could accurately 12 (AUCs 0.71–0.90). additive value was limited. Multiclass prediction groups showed considerable misclassification, subclinical differentiate between externalizing internalizing (AUC 0.79). Conclusions These results suggest that models can conversion during critical period childhood using symptoms. enable personalization preventative interventions for increased risk, which reduce incidence problems.

Language: Английский

---

Comparing the influence of social risk factors on machine learning model performance across racial and ethnic groups in home healthcare

Nursing Outlook, Journal Year: 2025, Volume and Issue: 73(3), P. 102431 - 102431

Published: May 1, 2025

Language: Английский

---

Longitudinal Prediction of Mental Health Outcomes in Vulnerable Youth using Machine Learning

Published: May 13, 2025

Mental illnesses affect almost 15% of the world's population, with half cases emerging before age 14. Improved methods for predicting progression mental distress among adolescents, particularly in vulnerable populations, are needed. This study utilized traditional machine learning techniques to predict health status at 17. We assessed correlates outcomes a sample 632 adolescents general (i.e., total difficulties score 17 or higher) 11, who participated UK Millennium Cohort Study. was best predicted using Balanced Random Forest model (AUC 0.75). Explainability enabled identification several critical factors, such as school environment, emotional distress, sleep patterns, patience, and social network ages 11 14, which were able differentiate participants poor good Individuals experiencing persistent between most likely suffer from unhappiness academic struggles. Our results point potentially modifiable factors associated high risk. These could pave way improved early intervention preventive strategies young people during adolescence.

Language: Английский

---

Antisocial behavior is associated with reduced frontoparietal network efficiency in youth

Social Cognitive and Affective Neuroscience, Journal Year: 2023, Volume and Issue: 18(1)

Published: Jan. 1, 2023

Youth antisocial behavior (AB) is associated with deficits in socioemotional processing, reward and threat processing executive functioning. These are thought to emerge from differences neural structure, functioning connectivity, particularly within the default, salience frontoparietal networks. However, relationship between AB organization of these networks remains unclear. To address this gap, current study applied unweighted, undirected graph analyses resting-state functional magnetic resonance imaging data a cohort 161 adolescents (95 female) enriched for exposure poverty, risk factor AB. As prior work indicates that callous-unemotional (CU) traits may moderate neurocognitive profile youth AB, we examined CU as moderator. Using multi-informant latent factors, was found be less efficient network topology, effect limited at low or mean levels traits, indicating were specific those high on but not traits. Neither nor their interaction significantly related default topologies. Results suggest specifically, linked shifts architecture network.

Language: Английский

---

Different whole-brain functional connectivity correlates of reactive-proactive aggression and callous-unemotional traits in children and adolescents with disruptive behaviors

NeuroImage Clinical, Journal Year: 2023, Volume and Issue: 40, P. 103542 - 103542

Published: Jan. 1, 2023

Disruptive behavior in children and adolescents can manifest as reactive aggression proactive is modulated by callous-unemotional traits other comorbidities. Neural correlates of these dimensions or subtypes comorbid symptoms remain largely unknown. This multi-center study investigated the relationship between resting state functional connectivity (rsFC) considering The large sample aged 8–18 years (n = 207; mean age 13.30 ± 2.60 years, 150 males) included 118 cases with disruptive (80 Oppositional Defiant Disorder and/or Conduct Disorder) 89 controls. Attention-deficit/hyperactivity disorder (ADHD) anxiety symptom scores were analyzed covariates when assessing group differences dimensional effects on hypothesis-free global local voxel-to-voxel whole-brain rsFC based magnetic resonance imaging at 3 Tesla. Compared to controls, demonstrated altered frontal areas, but not ADHD controlled. For cases, related central gyrus precuneus, regions linked aggression-related impairments. Callous-unemotional trait severity was correlated ICC inferior middle temporal implicated empathy, emotion, reward processing. Most observed subtype-specific patterns could only be identified controlled for. clarifies that brain measures disentangle distinct though overlapping youths. Moreover, our results highlight importance detect alterations

Language: Английский

---

Selectively predicting the onset of ADHD, oppositional defiant disorder, and conduct disorder in early adolescence with high accuracy

Frontiers in Psychiatry, Journal Year: 2023, Volume and Issue: 14

Published: Dec. 8, 2023

Introduction The externalizing disorders of attention deficit hyperactivity disorder (ADHD), oppositional defiant (ODD), and conduct (CD) are common in adolescence strong predictors adult psychopathology. While treatable, substantial diagnostic overlap complicates intervention planning. Understanding which factors predict the onset each disambiguating their different is translational interest. Materials methods We analyzed 5,777 multimodal candidate from children aged 9–10 years parents ABCD cohort to future ADHD, ODD, CD at 2-year follow-up. used deep learning optimized with an innovative AI algorithm jointly optimize model training, perform automated feature selection, construct individual-level predictions illness all prevailing cases 11–12 examined relative predictive performance when were restricted only neural metrics. Results Multimodal models achieved ~86–97% accuracy, 0.919–0.996 AUROC, ~82–97% precision recall testing held-out, unseen data. In neural-only models, dropped substantially but nonetheless accuracy AUROC ~80%. Parent aggressive traits uniquely differentiated while structural MRI metrics limbic system specific CD. Psychosocial measures sleep disorders, parent mental health behavioral traits, school proved valuable across disorders. functional subcortical regions cortical-subcortical connectivity emphasized. Overall, we identified a correlation between final predictor importance. Conclusion Deep can generate highly accurate early adolescent using features. frequently co-morbid adolescents, certain ODD or vs. ADHD. To our knowledge, this first machine study three major same design participant enable direct comparisons, analyze >200 features, include many types neuroimaging Future test observations external validation data will help further generalizability these findings.

Language: Английский

---

A review of information sources and analysis methods for data driven decision aids in child and adolescent mental health services

International Journal of Medical Informatics, Journal Year: 2024, Volume and Issue: 188, P. 105479 - 105479

Published: May 13, 2024

Clinical data analysis relies on effective methods and appropriate data. Recognizing distinctive clinical services service functions may lead to improved decision-making. Our first objective is categorize analytical methods, sources, algorithms used in current research information decision support child adolescent mental health (CAMHS). secondary identify the potential for different which data-driven aids can be useful.

Language: Английский

---