A Convolutional Neural Network Combined With Prototype Learning Framework for Brain Functional Network Classification...

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Medical Diagnostics

Background:

• Deep learning is increasingly used for brain disease diagnosis.
• Functional magnetic resonance imaging (fMRI) allows for the construction of brain functional networks.
• Autism Spectrum Disorder (ASD) diagnosis can benefit from advanced computational methods.

Purpose of the Study:

• To propose a novel deep learning framework, CNNPL, for classifying brain functional networks to aid in ASD diagnosis.
• To develop a robust method for identifying biomarkers associated with ASD from fMRI data.

Main Methods:

• Constructed brain functional networks using fMRI data.
• Developed a Convolutional Neural Network combined with Prototype Learning (CNNPL) framework.
• Employed a generalized prototype loss and transfer learning for model training and classification.
• Utilized prototype matching for classification based on learned features.

Main Results:

• The CNNPL model outperformed existing state-of-the-art methods in ASD classification on a multi-site dataset.
• The model demonstrated robustness in learning inter-site biomarkers despite data variability.
• Identified specific brain regions as reliable biomarkers for ASD classification.
• Showcased strong capability in learning high-level organization of brain functionality.

Conclusions:

• The proposed CNNPL framework offers a promising solution for learning and classifying brain functional networks.
• This approach contributes to biomarker extraction and imaging-based diagnosis of ASD.
• The model's robustness and performance highlight its potential for clinical application in neurological disorders.