چکیده:
Background: Artificial intelligence (AI) and deep learning (DL) have substantially advanced the analysis of brain imaging data, particularly in deciphering complex brain activity patterns from functional magnetic resonance imaging (fMRI). However, a consolidated evaluation of the efficacy of various DL algorithms in identifying and classifying patterns associated with autism spectrum disorder (ASD) in children is still needed. This umbrella review aims to systematically synthesize and compare evidence from existing systematic reviews (SRs) on the application of DL for detecting and classifying brain activity patterns in pediatric ASD using fMRI.
Materials and Methods: This review was conducted in accordance with the Preferred Reporting Items for Overviews of Reviews (PRIOR) guideline. A systematic search was performed across four major electronic databases (PubMed/Medline, Web of Science, Scopus, and Embase) using predefined keywords related to DL, fMRI, and ASD. Additionally, the methodological quality and risk of bias in the included systematic reviews were assessed using the Joanna Briggs Institute (JBI) critical appraisal tool.
Results: Seven systematic reviews, encompassing a total of 73 original studies, were included. The most prevalent DL architectures employed were Convolutional Neural Networks (CNNs) and Deep Belief Networks (DBNs). These models were applied for the detection and classification of ASD from fMRI data. The reported accuracy across the included studies ranged from 0.60 to 0.95, varying significantly based on the specific algorithm, dataset, and methodology used.
Conclusion: The evolution of DL algorithms for ASD detection and classification via fMRI represents a promising advancement. These models demonstrate a superior capacity to extract intricate features from neuroimaging data, achieving diagnostic accuracy that can rival or exceed human expertise. This technology holds substantial potential to revolutionize diagnostic protocols in clinical psychiatry and psychology, potentially leading to earlier intervention and improved patient outcomes.
Materials and Methods: This review was conducted in accordance with the Preferred Reporting Items for Overviews of Reviews (PRIOR) guideline. A systematic search was performed across four major electronic databases (PubMed/Medline, Web of Science, Scopus, and Embase) using predefined keywords related to DL, fMRI, and ASD. Additionally, the methodological quality and risk of bias in the included systematic reviews were assessed using the Joanna Briggs Institute (JBI) critical appraisal tool.
Results: Seven systematic reviews, encompassing a total of 73 original studies, were included. The most prevalent DL architectures employed were Convolutional Neural Networks (CNNs) and Deep Belief Networks (DBNs). These models were applied for the detection and classification of ASD from fMRI data. The reported accuracy across the included studies ranged from 0.60 to 0.95, varying significantly based on the specific algorithm, dataset, and methodology used.
Conclusion: The evolution of DL algorithms for ASD detection and classification via fMRI represents a promising advancement. These models demonstrate a superior capacity to extract intricate features from neuroimaging data, achieving diagnostic accuracy that can rival or exceed human expertise. This technology holds substantial potential to revolutionize diagnostic protocols in clinical psychiatry and psychology, potentially leading to earlier intervention and improved patient outcomes.
| بازنشر اطلاعات | |
 |
این مقاله تحت شرایط Creative Commons Attribution-NonCommercial 4.0 International License قابل بازنشر است. |
