•  
  •  
 

Corresponding Author

Hadeel M. Saleh

Authors ORCID

Yasir Khalaf Rashid: https://orcid.org/0000-0003-3210-4798

Taisir Abdulrazzaq Jar Alisawi: https://orcid.org/0009-0009-1748-1641

Tayseer Omar Hindy: https://orcid.org/0009-0003-1239-1838

Document Type

Article

Keywords

Autism spectrum disorder (ASD), Deep neural networks, True negative, False positive, Deep learning model

Abstract

Autism Spectrum Disorder (ASD) represents a complex neurological disorder characterized by difficulties in social engagement communication and repetitive behaviors. Identifying and intervening as soon as possible will result in improving the quality of life among children diagnosed with autism .In this study strives to provide lifestyle based relatively simple, practical and most importantly accurate deep learning autism diagnosis model of differential diagnosis of autism among children. Multiple deep-learning architectures such as deep neural networks (DNNs) were trained and compared. The proposed model attained outstanding performance, as reflected by the 100% classification ability of the model in the research results. We observed precision, recall, and F1 scores of 1.00 across all of the classes, indicating that the model achieved perfect classification performance on the dataset. The confusion matrix of the model also showed full True Positive and True Negative which means 0 False Positive or False Negative. The obtained results show that the proposed differential deep learning model is a reliable and state-of-the-art method for autism diagnosis among children, facilitating accurate and early diagnosis by medical practitioners as well as accelerating their intervention and treatment. Although the model achieved 100% accuracy on the test dataset, this may indicate potential overfitting due to dataset size or characteristics. Further validation using external datasets is recommended.

Download

Share

COinS