A Discrete Congruence Levenberg-Marquardt Deep Convoluted Neural Learning Classifier for the Automatic Detection of Autism Spectrum Disorder

Document Type

Article

Publication Title

Journal of Computational and Cognitive Engineering

Abstract

Autism spectrum disorder is a condition that affects around one out of every 54 children. Many studies have identified abnormalities in electroencephalography (EEG) signals for ASD diagnosis. The early and accurate identification of autism poses a substantial difficulty. The detection accuracy needs to be significantly boosted and the computational complexity reduced. The discrete congruence Levenberg-Marquardt deep convoluted neural learning classification (DCLMDCNLC) approach is introduced to address these issues in this work. The goal of the DCLMDCNLC approach is to perform automated ASD diagnosis at an early stage with higher accuracy and less time complexity. The DCLMDCNLC technique is applied to EEG signals through pre-processing, feature selection, and data classification. Discrete global threshold wavelet-transform-based pre-processing is carried out for EEG signal decomposition to remove unwanted noise. After that congruence correlation feature selection is carried out using the DCLMDCNLC technique with denoised signals to perform further processing. Finally, piecewise regression data analysis is carried out using the DCLMDCNLC technique for accurate autism detection with higher accuracy. An experimental assessment of the DCLMDCNLC technique is simulated, and the technique is validated using the EEG dataset for autism detection. Compared with traditional approaches, the DCLMDCNLC technique improves the accurate diagnosis of autism by 65%, the precision by 15%, the recall by 17%, the rate of errors by 77%, and the autism detection time by 35%.

First Page

36

Last Page

46

DOI

10.47852/bonviewJCCE42023620

Publication Date

2-21-2025

This document is currently not available here.

Share

COinS