A Matching Energy based approach in wavelet domain to distinguish hyper-responsive behavior in Autism Spectrum Disorders

Hyper-or hypo-responsive behavior to sensory stimulus confers heterogeneity to Autism Spectrum Disorders (ASD), and deteriorates the communication skills and learning ability of such individuals. Treatment should be tailored to the individual based on inter-individual variability. The current study is an attempt to differentiate the hyper-responsive subgroup from ASD by examining the complexity variations in Visual Evoked Potentials (VEPs), measured from a group of 30 individuals with ASD, in presence of achromatic sinusoidal gratings under low (5%), medium (50%) and high (90%) contrasts. This study proposes a diagnostic framework that includes an optimization-based denoising method, wavelet decomposition and extraction of matching energy (ME) to analyze complexity. Transient Artifact Reduction Algorithm (TARA) is employed to eliminate majority of the noise components in VEPs. Seven-level wavelet decomposition is performed on the denoised VEPs to decompose it into five frequency bands of gamma, beta, alpha, theta and delta. ME, a nonlinear complexity metric is extracted from these bands to perform relative complexity analysis (RC) with the variation in contrasts. The ME-based RC analysis reveals an increasing complexity pattern with progressing contrasts in 43%-50% of the ASD group in all the frequency bands. This pattern highlights the presence of heterogeneity and identifies the hyper-responsive subgroup in ASD. Overall, the results of this research highlight the potential of nonlinear dynamics in examining the heterogeneity and discerning the hyper-responsive subgroup in ASD thus aiding in targeted medical intervention which could improve the social skills of individuals with ASD.