Artificial intelligence for tracking social behaviours and supporting an autism spectrum disorder diagnosis: systematic review and meta-analysis.

IF 10.8 1区医学 Q1 MEDICINE, RESEARCH & EXPERIMENTAL

EBioMedicine Pub Date : 2025-09-26 DOI:10.1016/j.ebiom.2025.105931

Carter Sun, Alistair McEwan, Kelsie A Boulton, Eleni Andrea Demetriou, Ayesha K Sadozai, Amit Lampit, Adam J Guastella

{"title":"Artificial intelligence for tracking social behaviours and supporting an autism spectrum disorder diagnosis: systematic review and meta-analysis.","authors":"Carter Sun, Alistair McEwan, Kelsie A Boulton, Eleni Andrea Demetriou, Ayesha K Sadozai, Amit Lampit, Adam J Guastella","doi":"10.1016/j.ebiom.2025.105931","DOIUrl":null,"url":null,"abstract":"Background: Artificial intelligence (AI) holds promise for developing tools that can track social behaviours and support clinical assessments and outcomes in Autism Spectrum Disorders (ASD). This review evaluated existing AI algorithms for extracting facial information during social interaction assessments and contributing to diagnostic accuracy for ASD assessment and response to therapy.Methods: Systematic review of studies on human participants with an ASD diagnosis, sourced from Medline, Embase, Scopus, Web of Science, IEEE Xplore, and ACM Digital Library, evaluated the diagnostic accuracy of AI algorithms in ASD classification and their use in tracking social development through facial information for clinical application in social interactions. Bivariate and multi-level models addressed dependencies, heterogeneity, moderators (modalities, algorithms, tasks), and applied robust variance estimation. Publication bias was evaluated with funnel plots. The QUADAS-2 tool assessed the risk of bias and applicability. This study was registered on PROSPERO (CRD42021249905).Findings: Of 40,570 studies identified, 38 met the review criteria, and seven provided sufficient data for meta-analysis. The pooled diagnostic odds ratio of 15.917 (95% CI [4.775-53.059]), and bivariate analysis estimated an area under the receiver operating characteristic curve of 0.862. Accuracy improved with facial features, unstructured play, support vector machines, and decision tree-based algorithms. AI methods can analyse social behaviours, including eye gaze on social stimuli, emotional expression, and joint attention in ASD assessments. AI-enabled robots have also been used to guide therapy.Interpretation: This study shows that AI can accurately and objectively augment ASD assessments, track social behaviours, and enhance therapy outcomes. Further validation in diverse populations is needed to ensure clinical applicability and ethical use.Funding: None.","PeriodicalId":11494,"journal":{"name":"EBioMedicine","volume":"120 ","pages":"105931"},"PeriodicalIF":10.8000,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"EBioMedicine","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.1016/j.ebiom.2025.105931","RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"MEDICINE, RESEARCH & EXPERIMENTAL","Score":null,"Total":0}

引用次数: 0

Abstract

Background: Artificial intelligence (AI) holds promise for developing tools that can track social behaviours and support clinical assessments and outcomes in Autism Spectrum Disorders (ASD). This review evaluated existing AI algorithms for extracting facial information during social interaction assessments and contributing to diagnostic accuracy for ASD assessment and response to therapy.

Methods: Systematic review of studies on human participants with an ASD diagnosis, sourced from Medline, Embase, Scopus, Web of Science, IEEE Xplore, and ACM Digital Library, evaluated the diagnostic accuracy of AI algorithms in ASD classification and their use in tracking social development through facial information for clinical application in social interactions. Bivariate and multi-level models addressed dependencies, heterogeneity, moderators (modalities, algorithms, tasks), and applied robust variance estimation. Publication bias was evaluated with funnel plots. The QUADAS-2 tool assessed the risk of bias and applicability. This study was registered on PROSPERO (CRD42021249905).

Findings: Of 40,570 studies identified, 38 met the review criteria, and seven provided sufficient data for meta-analysis. The pooled diagnostic odds ratio of 15.917 (95% CI [4.775-53.059]), and bivariate analysis estimated an area under the receiver operating characteristic curve of 0.862. Accuracy improved with facial features, unstructured play, support vector machines, and decision tree-based algorithms. AI methods can analyse social behaviours, including eye gaze on social stimuli, emotional expression, and joint attention in ASD assessments. AI-enabled robots have also been used to guide therapy.

Interpretation: This study shows that AI can accurately and objectively augment ASD assessments, track social behaviours, and enhance therapy outcomes. Further validation in diverse populations is needed to ensure clinical applicability and ethical use.

Funding: None.

查看原文本刊更多论文

跟踪社会行为和支持自闭症谱系障碍诊断的人工智能：系统回顾和荟萃分析。

背景：人工智能（AI）有望开发出能够跟踪自闭症谱系障碍（ASD）的社会行为和支持临床评估和结果的工具。本综述评估了在社会互动评估中提取面部信息的现有人工智能算法，并有助于ASD评估的诊断准确性和对治疗的反应。方法：系统回顾来自Medline、Embase、Scopus、Web of Science、IEEE Xplore和ACM数字图书馆的ASD诊断研究，评估人工智能算法在ASD分类中的诊断准确性，以及它们在通过面部信息跟踪社会发展方面的应用，以用于临床社交应用。双变量和多级模型处理依赖性、异质性、调节因子（模式、算法、任务），并应用稳健方差估计。用漏斗图评价发表偏倚。QUADAS-2工具评估偏倚风险和适用性。本研究已在PROSPERO注册（CRD42021249905）。结果：在40,570项研究中，38项符合评价标准，7项为meta分析提供了足够的数据。合并诊断优势比为15.917 (95% CI[4.775-53.059])，双变量分析估计受试者工作特征曲线下面积为0.862。人脸特征、非结构化玩法、支持向量机和基于决策树的算法提高了准确率。人工智能方法可以分析社会行为，包括对社会刺激的注视、情绪表达和ASD评估中的共同注意。人工智能机器人也被用于指导治疗。本研究表明，人工智能可以准确客观地增强ASD评估，跟踪社会行为，并提高治疗效果。需要在不同人群中进一步验证，以确保临床适用性和伦理使用。资金:没有。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

EBioMedicine Biochemistry, Genetics and Molecular Biology-General Biochemistry,Genetics and Molecular Biology

CiteScore

17.70

自引率

0.90%

发文量

579

审稿时长

5 weeks

期刊介绍： eBioMedicine is a comprehensive biomedical research journal that covers a wide range of studies that are relevant to human health. Our focus is on original research that explores the fundamental factors influencing human health and disease, including the discovery of new therapeutic targets and treatments, the identification of biomarkers and diagnostic tools, and the investigation and modification of disease pathways and mechanisms. We welcome studies from any biomedical discipline that contribute to our understanding of disease and aim to improve human health.