Diagnosis of autism spectrum disorder using EEMD and multiscale fluctuation based dispersion entropy with Bayesian optimized light GBM

Being one of the most prominent neurological development disorder, autism spectrum disorder (ASD) has arisen as global issue. It emerges over the first few years of life and neurologist diagnose ASD by considering various factors. A biomarker, such as an electroencephalogram (EEG), may be employed to aid the diagnosis process as it can easily track abnormal changes in the patient’s brain. Therefore, this paper intends to propose an ensemble empirical mode decomposition based novel approach with multiscale fluctuation dispersion entropy for automated diagnosis of ASD patients using multichannel EEG signals. The various boosting classifiers which are empowered with Bayesian optimization based parameter estimation technique are considered to evaluate performance of proposed approach. Finally, the overall analysis of results exhibits that light gradient boosting machine (LGBM) produces maximum accuracy, specificity, and sensitivity of 99.59%, 99.37%, and 99.18% respectively with 0.9998 value of area under curve and 0.9919 value of Kappa statistic with optimally small computational time and minimum error measures.