Sales-Training-Inspired Optimization for Deep High-Order Principal Network in Autism Spectrum Disorder Classification

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by social communication deficits and repetitive behaviours. Diagnosing ASD early is difficult for healthcare professionals due to its diverse and intricate presentation. However, early detection is vital for enhancing outcomes and enabling the children to access targeted therapies that support the development of social and communication skills. Moreover, Classical models were time-consuming and resource-intensive, and they required lengthy assessments and specialized training. To bridge these complications, this research proposes a Sales Training-Based Optimization enabled Deep High-Order Principal Component Network (STBO_DHPCNet) for ASD classification using resting-state fMRI (rs-fMRI) brain images from 1114 subjects in the ABIDE dataset. First, gamma correction is applied to enhance the quality of the autism brain image. Next, the Region of Interest (ROI) extraction is performed. Afterwards, the nub region extraction is performed based on Sales Training Based Optimization (STBO). On the other hand, feature extraction is done based on an enhanced brain image. Finally, the classification of ASD is done by using DHPCNet, and it is trained using STBO. Here, DHPCNet is developed by incorporating the Deep High-Order Attention Neural Network (DHA-Net) and Principal Component Analysis Network (PCA-Net). Moreover, the evaluation results show that the DHPCNet gained an increased range of accuracy, sensitivity and specificity as 95.62%, 94.79%, and 95.86%.