Related Experiment Video
Updated: Jul 1, 2025
11:41
Automated Sholl Analysis of Digitized Neuronal Morphology at Multiple Scales
Published on: November 14, 2010
33.6K
Automated whole slide morphometry of sural nerve biopsy using machine learning
Daisuke Ono1,2, Honami Kawai1, Hiroya Kuwahara1
1Department of Neurology and Neurological Science, Graduate School of Medical and Dental Sciences, Tokyo Medical and Dental University, Tokyo, Japan.
Neuropathology and Applied Neurobiology
|March 6, 2024
Summary
This study introduces an AI application for automated sural nerve biopsy analysis, enabling detailed morphometric measurements of nerve fibres and myelin. The software aids in understanding peripheral neuropathies by providing objective data for clinicians and researchers.
Area of Science:
- Neurology
- Computational Pathology
- Biomedical Imaging
Background:
- Sural nerve biopsy morphometry is crucial for diagnosing peripheral neuropathies.
- Manual analysis is labor-intensive, limiting comprehensive assessment.
- Automated analysis of whole slide images (WSIs) is needed.
Purpose of the Study:
- Develop a machine learning application for automated morphometric analysis of sural nerve biopsies.
- Input whole slide images (WSIs) to perform quantitative analyses.
- Facilitate a deeper understanding of peripheral neuropathies.
Main Methods:
- Developed a Python library with three supervised learning models: nerve fascicle segmentation, myelinated fibre detection, and myelin sheath segmentation.
- Fine-tuned models using 86 toluidine blue-stained slides.
- Utilized whole slide images (WSIs) for analysis.
Main Results:
- Achieved high performance metrics: 0.861 mask AP for fascicle segmentation, 0.711 box AP for fibre detection, and 0.817 mIoU for myelin segmentation.
- Analyzed 323,298 nerve fibres and 782 fascicles across 70 WSIs.
- Identified distinct fibre populations and quantified differences in demyelination and vasculitis groups.
Conclusions:
- An open-source application for automated whole slide morphometry of sural nerve biopsies has been developed.
- The software is user-friendly for clinicians and pathologists without machine learning expertise.
- Facilitates data-driven analysis for improved understanding of neuropathies.
