Deep Learning Model For Autism Diagnosing

Abstract

The brain development and physical appearance of the face are both impacted by the neurologic disorder known as Autism Spectrum Disorder (ASD). Children with ASD exhibit different facial landmarks from children who normally grow with Typical Developing (TD), despite the fact that the disorder is thought to be inherited. When a child's behavioral traits and facial features are examined, the likelihood of an accurate diagnosis is highest. In recent years, deep convolutional neural networks widely used in image processing because they can effectively extract features from images and learn to recognize patterns, making them well-suited for tasks such as object detection, image segmentation, and classification. In this paper, we provide a deep learning CNN-based model for diagnosing autism that divides children into two categories depending on their facial features: possibly healthy and potentially unhealthy. The Tesorflow and Keras libraries are used by the suggested deep learning model to carry out feature extraction and picture classification. The dataset obtained from the Kaggle repository is used to train and evaluate the model. The dataset that was used to test this model consisted of 2,122 that were excluded from the original 2,940 images due to the quality and race. The testing of the proposed model results in an Area Under Curve (AUC) of 99.8% and an accuracy of 97.3%. This model proves its high diagnosis accuracy, ease of use, and fast decision.