Related Experiment Video
Updated: Sep 16, 2025
09:47
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
1.3K
Deep learning-based feature selection for detection of autism spectrum disorder.
Ibrahim Nafisah1, Nermine Mahmoud2, Ahmed A Ewees3
1Department of Statistics and Operations Research, College of Sciences, King Saud University, Riyadh, Saudi Arabia.
Frontiers in Artificial Intelligence
|July 10, 2025
Summary
This study introduces a novel deep learning method for Autism Spectrum Disorder (ASD) detection using resting-state fMRI data. The approach achieves high accuracy, offering a promising tool for clinical diagnosis.
Area of Science:
- Neuroscience
- Artificial Intelligence
- Medical Imaging
Background:
- Autism Spectrum Disorder (ASD) presents diagnostic challenges due to symptom heterogeneity.
- Resting-state functional MRI (rs-fMRI) shows potential for identifying neural signatures of ASD.
- Current neuroimaging analyses face limitations like high dimensionality and noise.
Purpose of the Study:
- To develop a novel, accurate, and clinically applicable deep learning model for ASD detection.
- To enhance feature selection from rs-fMRI data for improved ASD identification.
- To overcome limitations in current neuroimaging-based ASD diagnostic approaches.
Main Methods:
- A hybrid deep learning model combining Stacked Sparse Denoising Autoencoder (SSDAE) and Multi-Layer Perceptron (MLP) was utilized.
- rs-fMRI data from the ABIDE I dataset, preprocessed with CPAC, was analyzed.
- An optimized Hiking Optimization Algorithm (HOA) with DynamicOpposites Learning (DOL) and Double Attractors was employed for feature selection.
Main Results:
- The proposed model achieved an average accuracy of 0.735.
- Sensitivity was recorded at 0.765 and specificity at 0.752.
- Performance metrics surpassed existing state-of-the-art methods in ASD detection.
Conclusions:
- The hybrid deep learning approach demonstrates significant effectiveness for ASD detection.
- Enhanced feature selection improves the accuracy of neuroimaging-based ASD diagnostic models.
- This study offers a promising direction for developing more precise and clinically viable ASD detection tools.
