Classification Model for Autism Spectrum Disorder Individuals: Utilizing Facial Grid-Wise Emotion Features and Dual-Branch Visual Transformation

Autism spectrum disorder (ASD) poses challenges in social interaction and attention, often identifiable through facial emotions due to the absence of definitive medical tests. This study introduces the Dual-branch CNN-based visual transformation model (Db-CNN-VTM) for ASD identification. Pre-processing and geometric data augmentation enhance model robustness, while feature extraction via transformation-coded convolutional neural network (Tr-CNN) and feature fusion using a residual network refined classification. Optimization through the hunger game search optimization (HGSO) method determines hyperparameters. Comparative analysis with ResNet, AlexNet, CNN, and DNN models demonstrates Db-CNN-VTM's superior performance with accuracy 96.91%, precision 96.91%, recall 96.89%, F1-score 96.90%, specificity 96.89%, FPR 0.03%, kappa 93.80%, MCC 93.80%, and MSE 0.03%.