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

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging

Background:

• Autism Spectrum Disorder (ASD) presents diverse challenges for early diagnosis.
• Traditional diagnostic methods are time-consuming and require specialized expertise.
• Early detection of ASD is crucial for effective intervention and improved outcomes.

Purpose of the Study:

• To develop an automated and efficient method for Autism Spectrum Disorder classification.
• To enhance the accuracy and accessibility of early ASD detection using neuroimaging data.
• To introduce the Sales Training-Based Optimization enabled Deep High-Order Principal Component Network (STBO_DHPCNet) for ASD diagnosis.

Main Methods:

• Utilized resting-state fMRI (rs-fMRI) data from 1114 subjects in the ABIDE dataset.
• Applied gamma correction for image enhancement and Region of Interest (ROI) extraction.
• Implemented Sales Training Based Optimization (STBO) for nub region extraction and DHPCNet training.
• Developed DHPCNet by integrating Deep High-Order Attention Neural Network (DHA-Net) and Principal Component Analysis Network (PCA-Net).

Main Results:

• The STBO_DHPCNet model achieved high classification performance.
• Achieved accuracy of 95.62%, sensitivity of 94.79%, and specificity of 95.86% for ASD detection.
• Demonstrated the effectiveness of the proposed model in classifying ASD from rs-fMRI images.

Conclusions:

• The STBO_DHPCNet offers a promising, accurate, and efficient approach for early Autism Spectrum Disorder detection.
• This AI-driven method can aid healthcare professionals in diagnosing ASD more effectively.
• The study highlights the potential of advanced deep learning techniques in neurodevelopmental disorder research.