It is argued that future early ASD detection systems should be developed as clinician-supervised decision-support tools rather than autonomous diagnostic instruments.
Abstract
Early detection of autism spectrum disorder (ASD) in young children is essential for timely referral, developmental monitoring, and access to early intervention. However, conventional screening and diagnostic pathways often depend on parent-report instruments, episodic clinical observation, and specialist-administered assessments, which may delay identification during the first years of life. This scoping review maps the methodological landscape of early ASD detection from traditional clinical screening to multimodal artificial intelligence (AI). A structured literature search was conducted across major biomedical, psychological, and engineering databases for studies published between January 2010 and May 2026. After screening and eligibility assessment, 65 evidence sources were included in the qualitative synthesis, with additional methodological guidelines used to support reporting and appraisal. The reviewed evidence shows that early ASD detection is increasingly shifting from single-session clinical assessment toward multidimensional risk characterization. Clinical and behavioral screening tools remain the foundation of early identification, while eye tracking, video-based motor analysis, acoustic and vocal biomarkers, electroencephalography (EEG), functional near-infrared spectroscopy (fNIRS), and molecular or genomic indicators provide complementary information across different developmental windows. AI-based methods, including machine learning, deep learning, Transformer architectures, multimodal fusion strategies, and foundation-model-based representation learning, may improve the objective quantification of gaze, movement, vocalization, neural activity, and biological risk. Nevertheless, most AI-assisted systems remain limited by small and heterogeneous datasets, insufficient external validation, population bias, privacy concerns, computational burden, and limited interpretability. This review argues that future early ASD detection systems should be developed as clinician-supervised decision-support tools rather than autonomous diagnostic instruments. Clinically meaningful progress will require robust external validation, privacy-preserving deployment, age-appropriate risk stratification, and intrinsically interpretable architectures that align model outputs with developmental and clinical knowledge.
Artificial intelligence shows promise as a supportive tool for early screening, but current evidence supports its use as a complement to, rather than replacement for, clinical assessment.
Andrea Catalina Mahecha Ballesteros, Juanita Valeria García Bello, Eleaine Scarlet González Zuñiga et al.· Current Psychiatry Reports· 0 citations
A narrative review critically analyzes early signs of ASD in childhood and the clinical, developmental, family, educational, and health-system consequences of delayed diagnosis and concludes that earlier recognition should not depend exclusively on a definitive label.
A. Porto, Eduarda Bispo Cazerta, Ian Max Targino Guerreiro Braga et al.· International Health Science...· 0 citations
Delays in diagnostic confirmation remain common in young children with autism spectrum disorder (ASD). These delays are particularly concerning for children with severe symptoms and elevated support needs, for whom early identification is especially important. There is therefore a need for objective and feasible approaches to assist early identification prior to specialist evaluation. Eye-tracking is a non-invasive method for quantifying gaze-fixation patterns associated with ASD. The present study examined whether gaze-fixation indices derived from the Gazefinder eye-tracking system can identify a clinically defined severe ASD subgroup within a real-world clinical population. The analysis included 442 children aged 2-6 years referred to a child psychiatry outpatient clinic who underwent Gazefinder assessment. Based on Childhood Autism Rating Scale (CARS) scores, children were classified into a Severe ASD group (n = 42) and an Other group (Non-ASD and Mild-to-moderate ASD; n = 400). Gaze fixation rates on predefined regions of interest were compared, and discriminative performance was evaluated using receiver operating characteristic analyses. Children in the Severe ASD group exhibited reduced fixation on the mouth region in dynamic facial stimuli and reduced fixation on people relative to geometry. A composite criterion derived from four gaze-fixation indices yielded a sensitivity of 85.7% and a specificity of 55.3% for discriminating Severe ASD. These findings suggest that Gazefinder-based measures may provide adjunctive information to support screening and referral-related decision-making for clinically defined severe ASD in young children.
Yoshimasa Mamiya, Kenji J. Tsuchiya, Taiichi Katayama et al.· Scientific Reports· 0 citations
A comprehensive review of recent advancements in ASD research, with particular emphasis on neuroimaging, artificial intelligence (AI), and machine learning (ML)-based diagnostic approaches, highlights the growing potential of AI-driven tools for supporting early ASD diagnosis and emphasizes the need for standardized protocols, external validation, explainable AI, and clinically translatable frameworks.
Kuljeet Singh, Khushi Mogha, S. Moctar· Neurological Sciences· 0 citations
ABSTRACT Early autism identification in children aged 0–5 years is often discussed in terms of screening accuracy, yet consequential delay frequently occurs across the pathway from first concern to referral, diagnostic assessment, and support. This critical narrative review examines early autism identification as a pathway problem rather than as a single testing event. It synthesizes evidence on developmental surveillance, autism‐specific screening, parental and clinician concern, referral conversion, diagnostic waiting, service capacity, inequity, family burden, and pre‐diagnostic support. Screening tools can identify an elevated likelihood of autism and may accelerate diagnosis for some screen‐positive children, but they cannot confirm diagnosis or safely exclude autism when concern persists, and they do not compensate for failures in follow‐up, referral, assessment capacity, or support initiation. Recent evidence supports a cautious interpretation of universal autism screening because diagnostic stability, screening accuracy, and intervention benefit in screen‐detected children remain uncertain. Comparative evidence on the Modified Checklist for Autism in Toddlers, Revised with Follow‐Up suggests context‐dependent performance, including variable sensitivity, low or inconsistent positive predictive value, and age‐dependent accuracy. Multicultural surveillance and implementation studies indicate that adding tools without aligning workflow, language support, follow‐up systems, and service capacity may not improve pathway performance. The review argues that early autism identification should be evaluated through linked quality measures: response to concern, repeated surveillance following negative or ambiguous screening, referral completion, time to diagnostic assessment, support initiation before diagnostic closure, and equity of access. The clinical priority is not a perfect screening instrument in isolation, but faster, more coherent, and more equitable local pathways that translate concern into timely action.
R. Kurmashev, Mavile Karaieva· Pediatric Investigation· 0 citations
BACKGROUND
Early identification of autism spectrum disorder (ASD) is crucial for improving outcomes, yet diagnosis is often delayed. Primary care faces challenges, including the limited accuracy of existing tools and scarce access to specialist assessments. This prospective cohort study evaluated the feasibility of a novel, structured examiner-child interactive play protocol as an early behavioral screening tool for ASD.
METHODS
We recruited 260 toddlers aged 18-36 months. At baseline, all children participated in a structured interactive play session. Their behaviors (e.g., eye contact, turn-taking, social smiling, symbolic play) were coded from video recordings to quantify nine core indicators. After a 6-month follow-up, 217 children completed comprehensive clinical assessments. Core symptoms were evaluated using the Childhood Autism Rating Scale (CARS), and developmental level was assessed using the Bayley Scales of Infant and Toddler Development-Fourth Edition (Bayley-4). Based on clinical assessment results, participants were classified into an ASD group (n = 45) and a typically developing (TD) group (n = 165) for primary analysis. Spearman correlation analysis was used to examine the relationships between behavioral indicators, developmental levels, and symptom severity. Receiver operating characteristic (ROC) curve analysis assessed each indicator's discriminative validity for ASD.
RESULTS
During the structured interactive play, the ASD group performed significantly worse on all behavioral indicators than the TD group. Correlation analysis revealed that several behavioral indicators, including eye contact, turn-taking, and symbolic play, showed positive correlations with cognitive (rs = 0.30-0.51) and language ability (rs = 0.34-0.43) and negative correlations with symptom severity (rs=-0.53--0.36). ROC analysis identified that eye-contact metrics were the strongest discriminators. Specifically, eye contact frequency yielded an AUC of 0.915 (sensitivity 96.36%, specificity 73.33%), and total eye contact duration achieved an AUC of 0.899 (sensitivity 86.67%, specificity 84.44%).
CONCLUSION
Behavioral observation through examiner-child structured interactive play, particularly metrics related to eye contact, effectively differentiates toddlers with iASD. This objective, low-cost method shows potential as a primary screening tool for enhancing early ASD detection in primary care settings.
CLINICAL TRIAL NUMBER
Not applicable.
Dan Ai, Binyue Hu, Qiuhong Wei et al.· BMC Psychiatry· 0 citations