Advanced Deep Learning Framework for Diagnosing Autism Spectrum Disorder Through Facial Expression Analysis

Autism Spectrum Disorder is a neurodevelopmental disorder that presents with persistent deficits in social communication and social or emotional reciprocity. Timely intervention may result from an early and precise diagnosis of autism spectrum disorder (ASD). This study proposes a paradigm that uses deep learning systems to analyze facial expressions and emotions to classify children with autism spectrum disorder. The proposed framework utilizes a hybrid model combining Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks (CNN-BiLSTM) model to learn and perform classification tasks, To enhance image quality, input data undergoes pre-processing techniques, including histogram equalization, which improves contrast, and adaptive median filtering, which reduces noise while preserving essential details. Model optimization is a self-adaptive Black-Winged kite optimization method for hyperparameter tuning, ensuring that the model's parameters are optimal for improved classification accuracy. Performance measurement results such as Accuracy, recall, precision, specificity, and confusion matrix shown from evaluating the framework in ASD datasets obtained from the Kaggle repository are implemented in Python. This approach enhances the speed and precision of Autism Spectrum Disorder diagnosis accordingly and gives a very powerful instrument for real-time clinical application.