Classification of autism spectrum disorder based on sample entropy of spontaneous functional near infra -red spectroscopy signal

Objectives To assess the possibility of distinguishing autism spectrum disorder (ASD) based on the characteristic of spontaneous hemodynamic fluctuations and to explore the location of abnormality in the brain. Methods Using the sample entropy (SampEn) of functional near-infrared spectroscopy (fNIRS) from bilateral inferior frontal gyrus (IFG) and temporal cortex (TC) on 25 children with ASD and 22 typical development (TD) children, the pattern of mind-wandering was assessed. With the SampEn as feature variables, a machine learning classifier was applied to mark ASD and locate the abnormal area in the brain. Results The SampEn was generally lower for ASD than TD, indicating the fNIRS series from ASD was unstable, had low fluctuation, and high self-similarity. The classification between ASD and TD could reach 97.6% in accuracy. Conclusions The SampEn of fNIRS could accurately distinguish ASD. The abnormality in terms of the SampEn occurs more frequently in IFG than TC, and more frequently in the left than in the right hemisphere. Significance The results of this study may help to understand the cortical mechanism of ASD and provide a fNIRS-based diagnosis for ASD.