Spectral analysis of Electroretinography to differentiate autism spectrum disorder and attention deficit hyperactivity disorder.

autism spectrum disorder (ASD) and attention deficit hyperactivity disorder (ADHD) are both neurodevelopmental conditions that produce social interaction and executive functioning challenges but require very different therapeutic strategies. For that reason, early and accurate differentiation is important. However, their heterogeneity and overlap in symptoms make ASD and ADHD difficult to differentiate. The current diagnostic procedure to detect and distinguish ASD and ADHD is lengthy as it involves a comprehensive medical, developmental, and behavioral assessment. A more accessible and faster screening tool is needed to avoid delays in treatment. There is evidence that some retinal responses captured by the electroretinogram (ERG) are reduced in ASD subjects compared to neurotypicals whereas an opposite trend has been reported in ADHD, making ERG a promising tool for differentiating ASD and ADHD. However, previous ERG analyses based on amplitude and timing of ERG waves have exhibited limited success in differentiating ASD and ADHD. Recently, it has been found that time-varying spectral analysis of ERG allows for more accurate ASD detection compared to time-domain analysis. In this study, we evaluated the feasibility of differentiation of ASD and ADHD using features obtained by decomposing ERG using variable frequency complex demodulation (VFCDM). We used VFCDM features to train machine learning models and evaluated them using a subject independent validation approach. We achieved a maximum accuracy of 84% (87% sensitivity, 79% specificity), outperforming previous studies using ERG. Features from higher frequencies were found to be more important than features from lower frequencies. Clinical Relevance-This study establishes high frequency ERG information as a potential biomarker to differentiate ASD and ADHD.