Development of Hybrid Convolutional Neural Network and Radial Basis Function for Autism Spectrum Disorder Classification

Huey Chern, Lim (2024) Development of Hybrid Convolutional Neural Network and Radial Basis Function for Autism Spectrum Disorder Classification. Masters thesis, Faculty of Cognitive Science and Human Development.

PDF Lim Huey Chern MSc Thesis FCS Final_rev AASAH 1.pdf Download (2MB)

Abstract

Autism spectrum disorder (ASD) has become a common topic. The symptoms and heterogeneity of individuals with ASD change over time. In addition, the assessment using rs-fMRI data to classify ASD fails to attain the biomarker standards, and the obstacle still does not settle in the previous study. According to the Centers for Disease Control and Prevention (CDC) report, in 2021, the prevalence rate of ASD will increases to 1 in 44 children. Therefore, further methods for higher classification performance must be investigated. Deep learning has recently dramatically improved the cutting edge in a wide range of artificial intelligence jobs. Hence, this study proposed hybrid deep learning algorithms for ASD classification. Two algorithms merged: U-net neural network and Radial Basis Function (RBF) for medical image segmentation. Besides, the convolutional neural network (CNN) was implemented and combined with the RBF algorithm for the ASD classification. In this research, the classification of ASD achieved 94.79% accuracy, a better result compared with previous research. This study also illustrates the applicability of deep learning methods in neuroscience research projects. Future research should explore diverse algorithms, validating them across varied datasets with different hyperparameters to enhance ASD classification efficiency.

Actions (For repository members only: login required)

View Item