Morphometric analysis of peripheral myelinated nerve fibers through deep learning

Area of Science:

• Neuroscience
• Computational Biology
• Medical Imaging

Background:

• Neurological diseases often involve axonal loss or demyelination.
• Accurate quantification of myelin and axons is crucial for disease study.
• Manual segmentation of nerve images is time-consuming and labor-intensive.

Purpose of the Study:

• To develop a convolutional neural network (CNN)-based approach for automated segmentation of myelin and axons in mouse nerve cross-section images.
• To improve the efficiency and accuracy of morphometric analysis in neurological research.

Main Methods:

• Adapted the U-Net architecture for image segmentation.
• Trained the CNN model using manually segmented nerve images, including those with pathologies.
• Compared morphometric data from the CNN model with manual measurements.

Main Results:

• The CNN-based approach significantly reduced analysis time.
• Achieved excellent accuracy in quantifying axonal density and g-ratio.
• Minor manual refinement was still needed, with small variations observed in axon diameter and myelin thickness (within 9.5%).

Conclusions:

• The developed CNN approach offers greatly increased efficiency for myelin and axon quantification.
• The AI model shows strong potential for accelerating neurological disease research.
• Future studies will address minor limitations to further enhance accuracy.