Generalized Autism Spectrum Disorder Prediction Using Contextual Aware Learning

Autism Spectrum Disorder (ASD) is a developmental condition that generally becomes apparent in early childhood and continues across the lifespan. This study aims to enhance the efficiency of ASD detection by applying machine learning techniques. While Functional Magnetic Resonance Imaging (fMRI) is commonly utilized due to its precision, it often lacks the capacity to deliver detailed diagnostic insights. This research proposes a novel approach that employs a custom-designed Convolutional Neural Network (CNN) to interpret resting-state fMRI data for early ASD identification in children. The CNN includes convolutional, pooling, normalization, dropout, and fully connected layers, specifically adjusted for high-dimensional inputs. Rather than relying on complex models and extensive parameter tuning, this work emphasizes data-driven strategies by implementing deep contextual learning over ASD datasets. The model prioritizes training efficiency and accuracy through the use of two enhancement techniques: translation and noise addition. The resulting improvements in prediction performance mark a meaningful step forward in ASD diagnosis.