Early Detection of Autism in Young Children Using AI and IoT: A Systematic Review Focused on Rwanda
Purpose: The aim of this study is to critically examine and synthesise existing evidence on the application of Artificial Intelligence (AI) and Internet of Things (IoT) technologies for the early detection of autism spectrum disorder (ASD) in young children, with particular attention to their relevance, feasibility, and potential adaptation within the Rwandan context. Design/Methodology/Approach: A comprehensive literature search was conducted across major electronic databases, including PubMed/MEDLINE, Scopus, IEEE Xplore, Web of Science, ACM Digital Library, and Google Scholar. The search combined controlled vocabulary (e.g., MeSH terms) and free-text keywords related to Artificial Intelligence (AI), Machine Learning (ML), Internet of Things (IoT), wearable sensors, computer vision, eye tracking, electroencephalography (EEG), physiological monitoring, developmental screening, and multimodal ASD detection. 185 publications were used. Systematic reviews were appraised with AMSTAR-2. Quality ratings informed the interpretation of findings but did not form the basis for exclusion. Research Limitation: Restricting the search to peer-reviewed English-language publications may have excluded relevant non-English studies or grey literature, introducing potential language and publication bias. Marked heterogeneity in study designs, populations, sensor modalities, AI methods, and outcome definitions precluded formal meta-analysis and limited direct comparability of predictive performance across studies. Consequently, reported model performance may not generalise to diverse populations or healthcare environments, including Rwanda. Findings: AI and IoT technologies show considerable potential for supporting early identification of ASD‑related developmental differences. Physiological and sensor‑derived indicators include EEG patterns, heart‑rate variability, electrodermal activity, sleep characteristics, movement, and other autonomic measures. Computer vision, eye tracking, wearable sensors, smartphones, and digital questionnaires provide increasingly accessible data-collection mechanisms. Multimodal approaches that integrate behavioural, physiological, developmental, and visual information appear particularly promising, as they capture complementary characteristics associated with ASD. Practical Implication: Rwanda could explore AI‑ and IoT‑supported developmental screening to complement conventional child‑development surveillance. Nevertheless, systems developed in other countries should undergo local adaptation, clinical evaluation, and external validation using Rwandan data before implementation. Social Implication: A locally validated and culturally appropriate system could facilitate earlier referral, expand opportunities for early intervention, and improve access to developmental services for children and families across Rwanda. Implementation must include safeguards for privacy, informed parental consent, data security, equity, and responsible AI use. Originality/Value: It highlights the need for a Rwanda‑specific multimodal dataset and an explainable, affordable, privacy‑preserving, and externally validated AI‑IoT framework that complements rather than replaces professional developmental assessment and clinical diagnosis.