Electromagnetic Interaction Algorithm (EIA)-Based Feature Selection With Adaptive Kernel Attention Network (AKAttNet)...

Area of Science:

• Neuroscience and Artificial Intelligence
• Computational Psychiatry
• Biomedical Data Analysis

Background:

• Autism spectrum disorder (ASD) diagnosis relies on identifying cognitive, social, and behavioral patterns.
• Current diagnostic methods often lack accuracy, efficient feature selection, and computational speed.
• Early and accurate ASD detection is critical for timely intervention and improved outcomes.

Purpose of the Study:

• To develop an integrated computational framework for enhanced autism spectrum disorder (ASD) detection.
• To improve the accuracy and efficiency of ASD classification using advanced machine learning techniques.
• To address limitations in traditional diagnostic methods for ASD.

Main Methods:

• Integration of the electromagnetic interaction algorithm (EIA) for optimal feature selection.
• Application of an adaptive kernel attention network (AKAttNet) for ASD classification.
• Evaluation on four diverse autism spectrum disorder (ASD) datasets, comparing against traditional and deep learning models.

Main Results:

• The proposed EIA-AKAttNet model achieved high classification accuracy, ranging from 0.901 to 0.9827 across datasets.
• Demonstrated superior performance compared to conventional machine learning and existing deep learning methods.
• Showcased efficient feature dimensionality reduction by EIA, leading to lower computational time and enhanced generalizability.

Conclusions:

• The hybrid EIA-AKAttNet framework offers a practical and effective solution for early autism spectrum disorder (ASD) diagnosis.
• This approach enhances diagnostic accuracy while reducing computational overhead, showing promise for clinical application.
• The study underscores the potential of combining deep learning with optimization algorithms for reliable ASD screening systems.