In this paper, the proposed work tests the computer vision application to perform the skill and emotion assessment of children with Autism Spectrum Disorder (ASD) by extracting various bio-behaviors, human activities, child-therapist interactions, and joint pose estimations from the video-recorded interactive single-or two-person play-based intervention sessions. A comprehensive data set of 300 videos are amassed from ASD children engaged in social interaction and developed three novel deep learning-based computer vision models which are explained as follows: 1) activity comprehension to analyze child-play partner interactions (Activity Comprehension model); 2) an automatic joint attention recognition framework using pose, and 3) emotion and facial expression recognition. We tested models on children’s real-world unseen 68 videos captured from the clinic and public datasets. The activity comprehension model has an overall accuracy of 72.32%, the joint attention models have an accuracy of 97% for following eye gaze and 93.4% for hand pointing and the facial expression recognition model has an overall accuracy of 95.1%. The proposed models could extract activities and behaviors of interest from free-play and intervention session videos, empowering clinicians with data useful in diagnosis, assessment, treatment formulation, and monitoring of ASD children with limited supervision.

Methods: Thirty individuals with ASD (11-18years) and thirty Typically Developing (TD) individuals (11-16years) were recruited to perform the mental task. The participants were instructed to flip the shown geometric images mentally and mark their response on a scale. The task-related multivariate EEG activations were analyzed using multiplex temporal Visibility Graphs (VGs) to compute local and global brain network functional connectivity dynamics. Results: With cognitive load (0-back to the 2-back task), the behavioral performance (d’ index and Reaction Time) has reduced in ASD. The brain network has become more segregated and less integrated, reflecting more involvement of intra-regions over inter-regions in ASD. The frequent rerouting in hubs measured by Eigenvector Centrality (EC) indicated the progression of brain trajectory towards ASD. Overall, the neural mechanisms involving hyperactive response in frontal regions, frequent rewiring, and strengthened brain connectivity as a result of learning-induced performance reflected the adaptation to cognitive demands in ASD. Discussion: The correlation between complex graph measures and behavioral domain further reflected that neural metrics could predict the behavioral performance of the individuals in the task. In the future, modeling functional connectivity-based markers have the potential to reflect the brain trajectory alterations, which can detect ASD even before the behavioral manifestations become apparent.

Recent complex network analysis reflected the brain network as a modular network with small-world architecture in Autism Spectrum Disorder (ASD). Network hierarchy, which can provide important information to comment on brain networks, especially in ASD, has not yet been fully explored. The present work proposes a Weighted Hierarchical Complexity (WHC) metric to study network topology using the node degree concept. To do so, brain networks have been constructed using a visibility algorithm. To ensure proper mapping of network characteristics by the proposed metric, it is statistically compared to other network measures of brain connectivity related to integration, segregation and centrality. Further, for automated ASD classification, these network metrics were fed to explainable machine learning algorithms and the results revealed that brain regions tend to hierarchically coordinate in ASD, but the hierarchical architecture is attenuated after a few steps compared to networks in Typically Developing individuals (TDs). The value of WHC (0.55) reveals architecture up to three levels (four-degree nodes) with an abundance of 2-degree hubs in ASD indicating high intra-connectivity compared to TDs (WHC=0.78; four-level spread). The explainable Support Vector Machine (SVM)-classifier model highlighted the role of WHC in classifying ASD with 98.76% of accuracy. The graph-theory metrics ensured that weaker long-range connections and stronger intra-connections are markers of ASD. Thus, it becomes evident that whole-brain architecture can be characterised by a chain-like hierarchical modular structure representing atypical brain topology as in ASD.

Preliminary evidence has documented functional connectivity during the cognitive task in Autism Spectrum Disorder (ASD). However, evidence of effective neural connectivity with respect to informati...