ALERT: Atlas-Based Low Estimation Rank Tensor Approach to Detect Autism Spectrum Disorder.

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Distinct brain regions activate and interact in response to stimuli.
• Functional connectivity analysis of fMRI data aids in diagnosing neurological disorders like autism spectrum disorder (ASD).
• Current ASD diagnosis relies on time-consuming behavioral assessments.

Purpose of the Study:

• To develop an automated approach for constructing functional connectivity networks from fMRI data.
• To utilize a deep neural network for categorizing topological similarities in these networks for ASD detection.
• To evaluate the efficacy of the proposed method using the ABIDE dataset.

Main Methods:

• Functional connectivity networks were constructed using time-series data from fMRI.
• Correlation matrices were calculated to represent brain region interactions.
• A 2D convolutional deep neural network was employed to analyze functional connectivity matrices.
• Brain atlases were used for mapping brain regions, and a majority voting strategy was applied.

Main Results:

• The proposed method successfully constructed low-rank tensor approximations of functional connectivity matrices.
• The 2D convolutional deep neural network model demonstrated the ability to categorize topological similarities.
• Testing on the ABIDE dataset yielded an ASD detection accuracy of 84.79%.
• The accuracy is comparable to existing state-of-the-art techniques.

Conclusions:

• The developed approach provides an effective and automated method for ASD detection using fMRI data.
• This technique offers a potential for early, cost-effective, and comprehensive diagnosis of ASD.
• Functional connectivity network analysis combined with deep learning shows significant promise for neurological disorder diagnosis.