ARTIFICIAL NEURAL NETWORKS BASED-PREDICTION OF AUTISM SPECTRUM DISORDER

Authors : İlknur UCUZ, Ayla UZUN CİCEK

Pages : 78-82

View : 19 | Download : 15

Publication Date : 2020-12-31

Article Type : Research Paper

Abstract :Aim: Autism Spectrum Disorders insert ignore into journalissuearticles values(ASD); is one of the important neurodevelopmental disorders. This study aimed to perform artificial-intelligence-based modeling based on the prenatal-perinatal factors, family history, and developmental characteristics, which are emphasized as risk factors for ASD in the literature. Materials and Methods: The study was designed with a retrospective management and data from 136 children with ASD and 143 healthy children were included. Results: According to the findings of the MLP model, the five most important factors were the mean age of first words insert ignore into journalissuearticles values(months);, the mean age of head control insert ignore into journalissuearticles values(months);, the mean age of sitting without support insert ignore into journalissuearticles values(months);, history of autism in the family, and the mean paternal age at pregnancy insert ignore into journalissuearticles values(years);, respectively. Overall percentages of the training and testing samples were 91.4% and 88.0%. AUC for the model was 0.922 for the separation of the autism and control groups. Conclusion:The proposed model is able to successfully differentiate patients with autism spectrum disorders from healthy individuals and identify factors associated with the disease.
Keywords : artificial neural networks, autism, prenatal risk factors, perinatal risk factors