A Deep Learning Approach for Early Autism Detection

Abstract

Among the most widely studies neurodevelopmental condition characterized by major differences in social communication, interpersonal interaction, and behavioral patterns, with considerable variability across individuals. Early detection is critical for intervention and long-term developmental outcomes. However, conventional diagnostic procedures can be time consuming, subjective, and dependent on specialized professionals. In this work, we present a system based on deep learning for early detection of ASD for both facial images and video data and using a convolutional neural network to extract features from static images and video analysis to capture dynamic facial patterns that may reflect the behavior related to ASD. The system is built as a user friendly Streamlit web application that accepts image uploads or video and gives prediction results and downloadable reports. This system is particularly accessible and easy to use. This combination of static and dynamic facial analysis can capture high range of behavioral indicators with less human bias, which is well suited for preliminary screening in resource limited settings. There is an integrated chatbot that gives information about autism that raises awareness and engagement among the users. Artificial intelligence has a role to play in early screening and awareness for autism. This paper presents about detecting autism disorders using deep learning.