A novel network architecture combining central-peripheral deviation with image-based convolutional neural networks...

Area of Science:

• Neuroimaging
• Machine Learning
• Medical Image Analysis

Background:

• Brain imaging analysis, particularly with diffusion tensor imaging (DTI), presents challenges due to complex structures and subtle features.
• Advancements in magnetic resonance imaging (MRI) necessitate improved classification methods for enhanced clinical diagnosis.

Purpose of the Study:

• To propose a novel classification method for DTI data by integrating a convolutional neural network (CNN) with a multilayer neural network (MLP) utilizing central-peripheral deviation (CPD).
• To enhance the classification performance for DTI data, particularly in differentiating age groups.

Main Methods:

• A hybrid neural network architecture combining CNN for feature extraction and MLP with CPD for classification.
• CPD quantifies white matter diffusion dynamics by evaluating spatial deviations in diffusion coefficients.
• The method was applied to DTI datasets from traumatic brain injury (MagNeTS) and brain atlas (ICBM) studies.

Main Results:

• The proposed method demonstrated improvements in classification performance compared to existing CNN-based image classification techniques.
• Successful application to real DTI data from clinical studies, showing enhanced classification accuracy for two age groups.
• Reduced training loss and classification error were observed with the novel approach.

Conclusions:

• The novel neural network strategy effectively classifies DTI data by analyzing diffusion dynamics.
• This approach offers a promising advancement for improving diagnostic accuracy in neuroimaging research.
• The method shows potential for clinical applications in diagnosing conditions like traumatic brain injury.