Early Detection of Autism Spectrum Disorder (ASD)-A Deep Learning Approach

In this research, we put together a deep learning strategy that brings together the medical biomarker data from EEG tests and information from both clinical observations and inquiries to recognize the children with a Autism Spectrum Disorder (ASD). We expect that it can improve the effectiveness of identification significantly in addition to minimizing the expenditure. To begin with, our approach employed a pioneering tactic for selecting a neuronal signal landmarks derived from an eye movement, facial expression, and EEG results. This research showed that using a combination of Local Binary Pattern algorithm and deep learning classification led to a 87.50% accurate classification rate for multi-modal data merging. Finding a point to this method of detection as highly useful and effective for diagnosing ASD in young children. Graphs and confusion matrices unveiled different discrimination methods of EEG, personal behavior, and questionaries’ for discrepancy discrimination between ASD individuals and typically developing kids, with EEG being the most useful in drawing distinctions. Studies portray the physiological and behavioral important complementary characteristics. The combination of complementary information in this study proposed the deep learning approach has led to a significant improvement in classification accuracy.