Autism diagnosis with a CNN model based on genetic mutations

Abstract

Repetitive behaviors and social communication difficulties are hallmarks of autism spectrum disorders (ASD). Although the exact causes of this condition are yet unknown, a genetic component can be found in up to 25% of cases. It is preferable to identify ASD in children as early as possible because this allows for prompt interventions. The primary critical step to effective therapy and early mediation of influenced children is the recognizable proof of Autism based on objective pathogenic transformation screening. The hereditary scene of autism spectrum disorder (ASD) is diverse, encompassing a variety of genetic abnormalities involving nearly multiple genes (TSC, SHANK,…) with varying degrees of penetrance. In this work, we began by representing the A, T, C, and G sequence as the DNA gene causing autism and its alterations using scalogram images for classification. In order to categorize DNA sequences, we are now introducing a convolutional neural network model. Our model has demonstrated promissing performance in terms of learning and testing rates for classification, surpassing 88%.