Diagnosis of Autism Spectrum Disorders in Young Children Based on Resting-State Functional Magnetic Resonance Imaging...

Area of Science:

• Neuroscience and Artificial Intelligence
• Medical Imaging Analysis
• Developmental Disorders Research

Background:

• Rising global incidence of autism spectrum disorder (ASD).
• Critical need for early ASD diagnosis for effective treatment.
• Limitations of traditional clinical interviews and behavioral observations for early detection.

Purpose of the Study:

• To develop an intelligent model for diagnosing ASD in young children.
• To utilize resting-state functional magnetic resonance imaging (rs-fMRI) data.
• To employ convolutional neural networks (CNNs) for diagnostic classification.

Main Methods:

• Application of CNNs, a powerful deep learning algorithm, for ASD diagnosis.
• Implementation of 'combining classifiers' (dynamic and static approaches) and 'transfer learning' to address data limitations.
• Analysis of rs-fMRI data from young children (ages 5-10) from ABIDE I and ABIDE II datasets.

Main Results:

• The proposed model demonstrated superior accuracy, sensitivity, and specificity compared to previous studies on the ABIDE I dataset.
• Achieved best results on ABIDE I (accuracy=0.7273, sensitivity=0.712, specificity=0.7348) using Adamax optimization.
• Obtained acceptable classification results on ABIDE II (accuracy=0.700, sensitivity=0.582, specificity=0.804) and combined datasets (accuracy=0.7045, sensitivity=0.679, specificity=0.7421).

Conclusions:

• The developed CNN-based architecture is an efficient tool for early ASD diagnosis in young children.
• The model's ability to analyze rs-fMRI data offers a novel approach to diagnosing brain dysfunctions.
• This AI-driven method has the potential to significantly improve early detection rates for ASD.