Classification and Detection of Autism Spectrum Disorder Based on Deep Learning Algorithms.

Area of Science:

• Computational neuroscience
• Medical imaging analysis
• Machine learning applications in healthcare

Background:

• Autism spectrum disorder (ASD) is a neurological condition impacting brain development and physical features, including facial landmarks.
• Children with ASD exhibit distinct facial characteristics compared to typically developed individuals.
• Early detection of ASD is crucial for timely intervention and support.

Purpose of the Study:

• To design and develop a system for autism spectrum disorder detection using social media data and facial recognition.
• To leverage deep learning techniques for accurate identification of facial landmarks associated with ASD.
• To create an accessible web application for experimental ASD detection based on facial features.

Main Methods:

• Utilized deep learning, specifically convolutional neural networks (CNNs) with transfer learning.
• Employed pre-trained models: Xception, VGG19, and NASNETMobile for image classification.
• Tested models on a dataset of 2,940 face images from Kaggle, evaluating using accuracy, specificity, and sensitivity.

Main Results:

• The Xception model demonstrated the highest performance, achieving an accuracy of 91%.
• VGG19 and NASNETMobile models achieved accuracies of 80% and 78%, respectively.
• The developed system provides a practical tool for assessing facial features related to ASD.

Conclusions:

• Deep learning models, particularly Xception, show significant potential for detecting autism spectrum disorder through facial feature analysis.
• The developed web application offers a novel and accessible method for assisting communities and psychiatrists in experimental ASD detection.
• Facial landmark analysis using advanced AI techniques can serve as a valuable tool in the early identification of ASD.