Pain detection through facial expressions in children with autism using deep learning

Autism spectrum disorder (ASD) is a neurodevelopmental disease that manifests in atypical behavior, impaired social interaction, and communication difficulties. This neurological disorder is commonly diagnosed in early childhood and persists until adulthood. Individuals with autism often encounter challenges in areas such as social communication, repetitive behaviors, and sensory processing. Children diagnosed with ASD may encounter challenges in effectively expressing their discomfort, leading them to exhibit atypical or less obvious signs of pain. Consequently, it becomes imperative to identify and assess pain in these individuals, while also evaluating their medical status. This could potentially lead to suboptimal pain management and avoidable distress. Healthcare providers can enhance their ability to accurately diagnose and manage pain in children with autism by employing specialized artificial intelligence (AI) models specifically built to identify and evaluate pain-related symptoms. This research introduces a novel AI model that combines ResNeXt and Mediapipe techniques for the purpose of pain recognition in autistic children. The ResNeXt model has the capability to accurately classify facial expressions in real time. The face detection module provided by Mediapipe is utilized to detect faces within an image or video stream. The Mediapipe framework can be employed for the purpose of extracting facial landmarks, including the precise positions of the eyes, nose, and mouth, subsequent to the successful identification of the face. A convolutional neural network (CNN) is employed for the purpose of categorizing facial characteristics. The final model integrates the results of both ResNeXt + Mediapipe and CNN to accurately identify facial expressions and then assess pain levels in the face. The combination of ResNeXt and Mediapipe holds promise for the development of a face expression detection pipeline that exhibits high levels of accuracy, reliability, and efficiency.