AxonDeepSeg: automatic axon and myelin segmentation from microscopy data using convolutional neural networks.

Area of Science:

• Neuroscience
• Computational Biology
• Medical Imaging

Background:

• Accurate segmentation of axon and myelin in nervous system microscopy images is crucial for quantitative analysis of tissue microstructure.
• Existing segmentation algorithms often require manual parameter tuning and are not readily available.
• Developing automated and accurate segmentation methods is essential for advancing neuroscience research and validating imaging techniques.

Purpose of the Study:

• To introduce AxonDeepSeg, an open-source software utilizing deep learning for automated axon and myelin segmentation in microscopic images.
• To provide a user-friendly tool with an easy training procedure for custom model generation.
• To offer pre-trained models for scanning electron microscopy (SEM) and transmission electron microscopy (TEM) data.

Main Methods:

• Development of a convolutional neural network (CNN) architecture for segmentation.
• Implementation of a straightforward training pipeline for creating new segmentation models from labeled data.
• Training and validation of two distinct models using SEM and TEM datasets.

Main Results:

• Achieved high pixel-wise accuracy: 85% (rat SEM), 81% (human SEM), 95% (mice TEM), and 84% (macaque TEM).
• Successfully segmented a complete rat spinal cord slice, enabling extraction of morphological metrics.
• Demonstrated comparability of extracted metrics with existing literature findings.

Conclusions:

• AxonDeepSeg provides an accurate, automated, and open-source solution for axon and myelin segmentation in microscopy.
• The software facilitates quantitative analysis of nervous system microstructure across diverse species and imaging modalities.
• AxonDeepSeg is readily available, promoting wider adoption and reproducibility in neuroscience research.