Investigating the detection of autism spectrum disorders with electroencephalography and multi-input convolutional neural networks

Autism spectrum disorder (ASD) is defined as a neurodevelopmental disorder that begins in early childhood and is characterized by deficits in social communication and interaction, and the presence of limited and repetitive patterns of behavior and interest. Detection of autism spectrum disorders (ASD) using electroencephalography (EEG) and multi-input convolutional neural networks (CNNs) has become an important research topic in neuroscience and artificial intelligence in recent years. While EEG allows analyzing neurological conditions by measuring the electrical activity of the brain, ESAs are notable for their ability to extract meaningful features from complex data sets. Combining these methods holds promise for the early diagnosis of ASD and the development of individualized treatment approaches. In this study, we present a novel EEG-based approach using a multi-input one-dimensional Convolutional Neural Network (CNN) model. The study was conducted using data from a total of 29 children, 9 of whom were normal and 20 of whom were diagnosed with ASD. This dataset was obtained from King Abdulaziz University Hospital open access data sources. In experimental studies, various EEG channel combinations were tested, and accuracy values ranging from 80.36% to 89.06% were obtained. The highest accuracy rate, 89.06%, was achieved when the channels FZ, T3, C3, C4, CZ, PZ, and OZ were used as inputs to the system. These channels correspond to electrodes located along the midline from the nasal bridge to the occipital bone, as well as those positioned on the upper part of the head between the ears.