Detecção de autismo em imagens faciais usando redes neurais convolucionais

Abstract

Autism Spectrum Disorder (ASD) is a condition that, during childhood, manifests through cognitive and linguistic disorders. ASD poses a significant challenge for society and science due to various factors, such as its prevalence in the population and its presentation of subtle, complex, and enduring characteristics for patients' lives. Note that this is even more crucial in the case of children. Therefore, it is a challenge to correctly detect evidence that can lead to an early and reliable diagnosis, enabling families and health services to conduct more effective, inclusive therapies with a greater potential for success. Literature reports various multidisciplinary efforts to detect the syndrome, with the diagnosis being clinically conducted. This study aims to contribute to the detection of autism in children by investigating the practical potential of Convolutional Neural Networks (CNNs) applied to patients' facial images. The experimental results indicate that it is possible to achieve an accuracy above 82% by combining a CNN with face detection and alignment techniques.