Empirical Analysis of Nature-Inspired Algorithms for Autism Spectrum Disorder Detection Using 3D Video Dataset

TL;DR

This study addresses ASD detection from a 3D gait video dataset by integrating correlation-based feature ranking with nature-inspired optimization to perform efficient feature extraction, followed by supervised ML classification. The approach evaluates multiple combinations of ranking methods, optimization algorithms, and classifiers, reporting a top accuracy of $100 ext%$ using a Random Forest classifier with the Gravitational Search Algorithm for feature selection (380 features), and a highly compact alternative with $96.875 ext%$ accuracy using a CS-based feature selection (4 features). The method demonstrates substantial reductions in feature dimensionality and computational time while maintaining or improving performance, highlighting its potential for practical ASD screening from video-derived gait data. The work emphasizes robustness via 10-fold cross-validation and suggests future expansion to larger,More diverse datasets and integration with deep learning or real-time computer vision applications. Overall, the paper provides a data-driven, scalable framework for ASD diagnosis using 3D gait analysis and nature-inspired feature selection, with clear implications for clinical deployment and further research.

Abstract

Autism Spectrum Disorder (ASD) is a chronic neurodevelopmental disorder symptoms of which includes repetitive behaviour and lack of social and communication skills. Even though these symptoms can be seen very clearly in social but a large number of individuals with ASD remain undiagnosed. In this paper, we worked on a methodology for the detection of ASD from a 3-dimensional walking video dataset, utilizing supervised machine learning (ML) classification algorithms and nature-inspired optimization algorithms for feature extraction from the dataset. The proposed methodology involves the classification of ASD using a supervised ML classification algorithm and extracting important and relevant features from the dataset using nature-inspired optimization algorithms. We also included the ranking coefficients to find the initial leading particle. This selection of particle significantly reduces the computation time and hence, improves the total efficiency and accuracy for ASD detection. To evaluate the efficiency of the proposed methodology, we deployed various combinationsalgorithms of classification algorithm and nature-inspired algorithms resulting in an outstanding classification accuracy of $100\%$ using the random forest classification algorithm and gravitational search algorithm for feature selection. The application of the proposed methodology with different datasets would enhance the robustness and generalizability of the proposed methodology. Due to high accuracy and less total computation time, the proposed methodology will offer a significant contribution to the medical and academic fields, providing a foundation for future research and advancements in ASD diagnosis.

Figures (8)

• Figure 1: Organization of the paper
• Figure 2: Methodology adopted
• Figure 3: 3D Videos of children walking towards camera
• Figure 4: Accuracy Analysis of various combinations of feature extraction and classification algorithms
• Figure 5: Feature Analysis for Pearson Correlation Coefficient