The assessment of autism spectrum disorder (ASD) is based on semi-structured procedures addressed to children and caregivers. Such methods rely on the evaluation of behavioural symptoms rather than on the objective evaluation of psychophysiological underpinnings. Advances in research provided evidence of modern procedures for the early assessment of ASD, involving both machine learning (ML) techniques and biomarkers, as eye movements (EM) towards social stimuli. This systematic review provides a comprehensive discussion of 11 papers regarding the early assessment of ASD based on ML techniques and children’s social visual attention (SVA). Evidences suggest ML as a relevant technique for the early assessment of ASD, which might represent a valid biomarker-based procedure to objectively make diagnosis. Limitations and future directions are discussed.

Background: Adaptive systems serve to personalize interventions or training based on the user’s needs and performance. The adaptation techniques rely on an underlying engine responsible for processing incoming data and generating tailored responses. Adaptive virtual reality (VR) systems have proven to be efficient in data monitoring and manipulation, as well as in their ability to transfer learning outcomes to the real world. In recent years, there has been significant interest in applying these systems to improve deficits associated with autism spectrum disorder (ASD). This is driven by the heterogeneity of symptoms among the population affected, highlighting the need for early customized interventions that target each individual’s specific symptom configuration. Objective: Recognizing these technology-driven therapeutic tools as efficient solutions, this systematic review aims to explore the application of adaptive VR systems in interventions for young individuals with ASD. Methods: An extensive search was conducted across 3 different databases—PubMed Central, Scopus, and Web of Science—to identify relevant studies from approximately the past decade. Each author independently screened the included studies to assess the risk of bias. Studies satisfying the following inclusion criteria were selected: (1) the experimental tasks were delivered via a VR system, (2) system adaptation was automated, (3) the VR system was designed for intervention or training of ASD symptoms, (4) participants’ ages ranged from 6 to 19 years, (5) the sample included at least 1 group with ASD, and (6) the adaptation strategy was thoroughly explained. Relevant information extracted from the studies included the sample size and mean age, the study’s objectives, the skill trained, the implemented device, the adaptive strategy used, the engine techniques, and the signal used to adapt the systems. Results: Overall, a total of 10 articles were included, involving 129 participants, 76% of whom had ASD. The studies included level switching (7/10, 70%), adaptive feedback strategies (9/10, 90%), and weighing the choice between a machine learning (ML) adaptive engine (3/10, 30%) and a non-ML adaptive engine (8/10, 80%). Adaptation signals ranged from explicit behavioral indicators (6/10, 60%), such as task performance, to implicit biosignals, such as motor movements, eye gaze, speech, and peripheral physiological responses (7/10, 70%). Conclusions: The findings reveal promising trends in the field, suggesting that automated VR systems leveraging real-time progression level switching and verbal feedback driven by non-ML techniques using explicit or, better yet, implicit signal processing have the potential to enhance interventions for young individuals with ASD. The limitations discussed mainly stem from the fact that no technological or automated tools were used to handle data, potentially introducing bias due to human error.

Many symptoms of the autism spectrum disorder (ASD) are evident in early infancy, but ASD is usually diagnosed much later by procedures lacking objective measurements. It is necessary to anticipate the identification of ASD by improving the objectivity of the procedure and the use of ecological settings. In this context, atypical motor skills are reaching consensus as a promising ASD biomarker, regardless of the level of symptom severity. This study aimed to assess differences in the whole-body motor skills between 20 children with ASD and 20 children with typical development during the execution of three tasks resembling regular activities presented in virtual reality. The virtual tasks asked to perform precise and goal-directed actions with different limbs vary in their degree of freedom of movement. Parametric and non-parametric statistical methods were applied to analyze differences in children’s motor skills. The findings endorsed the hypothesis that when particular goal-directed movements are required, the type of action could modulate the presence of motor abnormalities in ASD. In particular, the ASD motor abnormalities emerged in the task requiring to take with the upper limbs goal-directed actions with low degree of freedom. The motor abnormalities covered (1) the body part mainly involved in the action, and (2) further body parts not directly involved in the movement. Findings were discussed against the background of atypical prospective control of movements and visuomotor discoordination in ASD. These findings contribute to advance the understanding of motor skills in ASD while deepening ecological and objective assessment procedures based on VR.

Clinical assessment procedures encounter challenges in terms of objectivity because they rely on subjective data. Computational psychiatry proposes overcoming this limitation by introducing biosignal-based assessments able to detect clinical biomarkers, while virtual reality (VR) can offer ecological settings for measurement. Autism spectrum disorder (ASD) is a neurodevelopmental disorder where many biosignals have been tested to improve assessment procedures. However, in ASD research there is a lack of studies systematically comparing biosignals for the automatic classification of ASD when recorded simultaneously in ecological settings, and comparisons among previous studies are challenging due to methodological inconsistencies. In this study, we examined a VR screening tool consisting of four virtual scenes, and we compared machine learning models based on implicit (motor skills and eye movements) and explicit (behavioral responses) biosignals. Machine learning models were developed for each biosignal within the virtual scenes and then combined into a final model per biosignal. A linear support vector classifier with recursive feature elimination was used and tested using nested cross-validation. The final model based on motor skills exhibited the highest robustness in identifying ASD, achieving an AUC of 0.89 (SD = 0.08). The best behavioral model showed an AUC of 0.80, while further research is needed for the eye-movement models due to limitations with the eye-tracking glasses. These findings highlight the potential of motor skills in enhancing objectivity and reliability in the early assessment of ASD compared to other biosignals.