The Application of Artificial Intelligence to Ecological Momentary Assessment Data in Suicide Research: A Systematic Review (Preprint) DOI Creative Commons
Ruth Melia, Katherine Musacchio Schafer, Megan L. Rogers

et al.

Journal of Medical Internet Research, Journal Year: 2024, Volume and Issue: unknown

Published: June 27, 2024

Ecological momentary assessment (EMA) captures dynamic processes suitable to the study of suicidal ideation and behaviors. Artificial intelligence (AI) has increasingly been applied EMA data in processes. This review aims (1) synthesize empirical research applying AI strategies behaviors; (2) identify methodologies collection procedures used, suicide outcomes studied, applied, results reported; (3) develop a standardized reporting framework for researchers future. PsycINFO, PubMed, Scopus, Embase were searched published articles investigation outcomes. The PRISMA (Preferred Reporting Items Systematic Reviews Meta-Analyses) guidelines used studies while minimizing bias. Quality appraisal was performed using CREMAS (adapted STROBE [Strengthening Observational Studies Epidemiology] Checklist Momentary Assessment Studies). In total, 1201 records identified across databases. After full-text review, 12 (1%) articles, comprising 4398 participants, included. application predict ideation, reported mean area under curve (0.74-0.86), sensitivity (0.64-0.81), specificity (0.73-0.86), positive predictive values (0.72-0.77). met between 4 13 16 recommended standards, with an average 7 items studies. poorly training treatment missing data. Findings indicate promise self-report prediction near-term ideation. within is burgeoning hampered by variations procedures. development adapted team address this. Open Science Framework (OSF); https://doi.org/10.17605/OSF.IO/NZWUJ PROSPERO CRD42023440218; https://www.crd.york.ac.uk/PROSPERO/view/CRD42023440218.

Language: Английский

Learning Robust and Sparse Principal Components with the α-Divergence DOI
Aref Miri Rekavandi, Abd‐Krim Seghouane, Robin J. Evans

et al.

IEEE Transactions on Image Processing, Journal Year: 2024, Volume and Issue: 33, P. 3441 - 3455

Published: Jan. 1, 2024

In this paper, novel robust principal component analysis (RPCA) methods are proposed to exploit the local structure of datasets. The derived by minimizing α-divergence between sample distribution and Gaussian density model. is used in different frameworks represent variants RPCA approaches including orthogonal, non-orthogonal, sparse methods. We show that classical PCA a special case our where reduced Kullback-Leibler (KL) divergence. It shown simulations recover underlying components (PCs) down-weighting importance structured unstructured outliers. Furthermore, using simulated data, it can be applied fMRI signal recovery Foreground-Background (FB) separation video analysis. Results on real world problems FB as well image reconstruction also provided.

Language: Английский

Citations

2
Risk factors for suicide in patients with colorectal cancer: A Surveillance, Epidemiology, and End Results database analysis DOI
Justin Dourado, Sameh Hany Emile, Anjelli Wignakumar

et al.

Surgery, Journal Year: 2024, Volume and Issue: unknown

Published: Oct. 1, 2024

Language: Английский

Citations

2
The Application of Artificial Intelligence to Ecological Momentary Assessment Data in Suicide Research: A Systematic Review (Preprint) DOI
Ruth Melia, Katherine Musacchio Schafer, Megan L. Rogers

et al.

Published: June 27, 2024

BACKGROUND Ecological Momentary Assessment (EMA) can capture highly dynamic processes and intense variability patterns suitable to the study of suicidal ideation behaviors. Artificial Intelligence (AI), in particular Machine Learning (ML) strategies, have increasingly been applied EMA data suicide research. OBJECTIVE The review aims (1) synthesize empirical research applying AI strategies behaviors, (2) identify methodologies used, collection procedures employed, outcomes studied, applied, results reported, (3) develop a standardized reporting framework for researchers future. METHODS PsycINFO, PubMed, SCOPUS EMBASE were searched articles published until June 2024. Studies that investigation (suicidal ideation, attempt, death), collected across devices (Smartphone, Personal Digital Assistant, PC, tablet) settings (clinical, community), included. Preferred Reporting Items Systematic Reviews Meta Analyses (PRISMA) guidelines used relevant studies while minimizing bias. Specific reported included sampling method, monitoring period, prompt latency, compliance, attrition, treatment missing data. Quality appraisal was performed using an adapted checklist (CREMAS). RESULTS 1,201 records identified databases. After full text review, 12 articles, comprising 4398 participants conducted psychiatric hospitals (n = 5), emergency departments 2), outpatient clinics medical residency programs 1), university mental health with some settings. Design features (sampling strategy, prompting frequency, response device data) varied studies. In application predict mean AUCs (0.74 0.86), sensitivity (0.64 0.81), specificity (0.73 positive predictive values (0.72 0.77). CONCLUSIONS within is small but burgeoning area high heterogeneity apparent standards. Findings indicate promise ML self-report prediction near-term ideation. development by team standardize on going forward. CLINICALTRIAL PROSPERO: CRD42023440218 Open Science Framework: https://doi.org/10.17605/OSF.IO/NZWUJ

Language: Английский

Citations

0
The Association between Suicidal Ideation and Subtypes of Comorbid Insomnia Disorder in Apneic Individuals DOI Open Access
Matthieu Hein, Benjamin Wacquier,

Matteo Conenna

et al.

Journal of Clinical Medicine, Journal Year: 2024, Volume and Issue: 13(19), P. 5907 - 5907

Published: Oct. 3, 2024

: Given the existence of higher suicidality in apneic individuals, this study aimed to determine potential role played by subtypes comorbid insomnia disorder (CID) occurrence suicidal ideation for specific subpopulation.

Language: Английский

Citations

0
Suicidality Prediction in Youth Crisis Text Line Users: Development and Validation of an Explainable Artificial Intelligence Text Classifier (Preprint) DOI Creative Commons
Julia Thomas, Antonia Lucht, Jacob Segler

et al.

JMIR Public Health and Surveillance, Journal Year: 2024, Volume and Issue: 11, P. e63809 - e63809

Published: Nov. 7, 2024

Background Suicide represents a critical public health concern, and machine learning (ML) models offer the potential for identifying at-risk individuals. Recent studies using benchmark datasets real-world social media data have demonstrated capability of pretrained large language in predicting suicidal ideation behaviors (SIB) speech text. Objective This study aimed to (1) develop implement ML methods SIBs crisis helpline dataset, transformer-based as foundation; (2) evaluate, cross-validate, model against traditional text classification approaches; (3) train an explainable highlight relevant risk-associated features. Methods We analyzed chat protocols from adolescents young adults (aged 14-25 years) seeking assistance German helpline. An was developed architecture with weights long short-term memory layers. The predicted (SI) advanced engagement (ASE), indicated by composite Columbia-Suicide Severity Rating Scale scores. compared performance classical word-vector-based model. subsequently computed discrimination, calibration, clinical utility, explainability information Shapley Additive Explanations value-based post hoc estimation Results dataset comprised 1348 help-seeking encounters (1011 training 337 testing). classifier achieved macroaveraged area under curve (AUC) receiver operating characteristic (ROC) 0.89 (95% CI 0.81-0.91) overall accuracy 0.79 0.73-0.99). surpassed baseline (AUC-ROC=0.77, 95% 0.64-0.90; accuracy=0.61, 0.61-0.80). transformer excellent prediction nonsuicidal sessions (AUC-ROC=0.96, 0.96-0.99) good SI ASE, AUC-ROCs 0.85 0.97-0.86) 0.87 0.81-0.88), respectively. Brier Skill Score 44% improvement over identified features predictive SIBs, including self-reference, negation, expressions low self-esteem, absolutist language. Conclusions Neural networks model–based transfer can accurately identify ASE. explainer revealed associated Such may potentially support decision-making suicide prevention services. Future research should explore multimodal input temporal aspects risk.

Language: Английский

Citations

0
The Application of Artificial Intelligence to Ecological Momentary Assessment Data in Suicide Research: A Systematic Review (Preprint) DOI Creative Commons
Ruth Melia, Katherine Musacchio Schafer, Megan L. Rogers

et al.

Journal of Medical Internet Research, Journal Year: 2024, Volume and Issue: unknown

Published: June 27, 2024

Ecological momentary assessment (EMA) captures dynamic processes suitable to the study of suicidal ideation and behaviors. Artificial intelligence (AI) has increasingly been applied EMA data in processes. This review aims (1) synthesize empirical research applying AI strategies behaviors; (2) identify methodologies collection procedures used, suicide outcomes studied, applied, results reported; (3) develop a standardized reporting framework for researchers future. PsycINFO, PubMed, Scopus, Embase were searched published articles investigation outcomes. The PRISMA (Preferred Reporting Items Systematic Reviews Meta-Analyses) guidelines used studies while minimizing bias. Quality appraisal was performed using CREMAS (adapted STROBE [Strengthening Observational Studies Epidemiology] Checklist Momentary Assessment Studies). In total, 1201 records identified across databases. After full-text review, 12 (1%) articles, comprising 4398 participants, included. application predict ideation, reported mean area under curve (0.74-0.86), sensitivity (0.64-0.81), specificity (0.73-0.86), positive predictive values (0.72-0.77). met between 4 13 16 recommended standards, with an average 7 items studies. poorly training treatment missing data. Findings indicate promise self-report prediction near-term ideation. within is burgeoning hampered by variations procedures. development adapted team address this. Open Science Framework (OSF); https://doi.org/10.17605/OSF.IO/NZWUJ PROSPERO CRD42023440218; https://www.crd.york.ac.uk/PROSPERO/view/CRD42023440218.

Language: Английский

Citations

0