A Multimodal AI Framework for Early Autism Identification Using CbRNN-Ensemble Modelling and Advanced NCIAPPF-Based...

Area of Science:

• Artificial Intelligence
• Neuroscience
• Developmental Pediatrics

Background:

• Early identification of autism spectrum disorder (ASD) is critical for timely interventions and improved developmental outcomes.
• Conventional diagnostic methods for ASD face limitations in objectivity, scalability, and reliance on expert assessment.
• There is a need for advanced, reliable, and scalable tools for early ASD detection.

Purpose of the Study:

• To introduce AutiScan, an AI-driven framework designed for early screening of autism spectrum disorder.
• To leverage a Convolution-based Recurrent Neural Network Ensemble Model (CbRNN-EM) integrated with NCIAPPF for analyzing temporal dynamics in neural encoding.
• To assess the efficacy of AutiScan in analyzing multimodal data for accurate ASD marker identification.

Main Methods:

• Development of AutiScan, an AI framework utilizing a CbRNN-EM and NCIAPPF technique.
• Analysis of multimodal data, including facial expressions, eye-gaze, speech prosody, and behavioral cues.
• Application of NCIAPPF for adaptive noise filtering, signal normalization, and preservation of temporal dependencies in multimodal data.

Main Results:

• AutiScan effectively extracts spatial, emotional, and sequential features relevant to early ASD markers from multimodal data.
• The NCIAPPF technique enhances preprocessing by filtering noise and preserving crucial temporal dynamics.
• The CbRNN-EM ensemble model achieved superior classification accuracy for ASD detection compared to traditional methods.

Conclusions:

• AutiScan provides a dependable, scalable, and interpretable tool for early autism spectrum disorder screening.
• The AI-driven approach demonstrates significant potential for practical implementation in clinical and educational settings.
• AutiScan facilitates prompt interventions by enabling earlier and more accurate identification of ASD.