HYBRID OPTIMIZATION-ENABLED FUNCTIONAL CONNECTIVITY-BASED PIVOTAL REGION EXTRACTION AND TRANSFER LEARNING FOR AUTISM SPECTRUM DISORDER DETECTION

Abstract

Autism Spectrum Disorder (ASD) is a neuro development-based disability caused by variations in the brain which occur during the early stages of human life. This may cause impact on social skills and communication of an individual. Autism is a highly challenging issue to diagnose at the early stages. ASD is one of the important problems to diagnose because it starts manifesting at low ages. However, diagnosing ASD is difficult due to its complex symptoms as well as an inadequate number of neurobiological indications. This paper aims to detect ASD based on pivotal region extraction and Feedback-Henry Jellyfish Optimization-Convolution Neural Network with Transfer Learning (FHJO_CNN with TL) is developed in this work. Here, an autistic brain image is acquired from the data sample and it is fed to the pre-processing stage. By utilizing the Kalman filter and Region of Interest (RoI), selection pre-processing is done. After that, functional connectivity-based pivotal region extraction is performed based on the proposed FHJO, where the FHJO is an incorporation of Feedback-Henry Gas Optimization (FHGO) and Jelly Fish (JS) algorithms, where, FHGO is the combination of Henry Gas Solubility Optimization (HGSO), and Feedback Artificial Tree (FAT) algorithm. Thereafter in feature extraction, features like Local Binary Pattern (LBP), and Speeded Up Robust Features (SURF), statistical features are excerpted. Finally, ASD classification is done utilizing the proposed FHJO_CNN with TL. The performance of the proposed method is evaluated using the ACERTA-ABIDE dataset and the evaluation reveals that FHJO_CNN with TL attained values of accuracy, sensitivity and specificity which are 94.08%, 95.09%, and 93.59% respectively.