Related Experiment Video
Updated: Oct 16, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.0K
Automated Segmentation of Median Nerve in Dynamic Sonography Using Deep Learning: Evaluation of Model Performance.
Chueh-Hung Wu1,2, Wei-Ting Syu3, Meng-Ting Lin1
1Department of Physical Medicine and Rehabilitation, National Taiwan University Hospital Hsin-Chu Branch, Hsinchu 300, Taiwan.
Diagnostics (Basel, Switzerland)
|October 23, 2021
Summary
Automated deep learning models accurately segment the median nerve in dynamic ultrasound for carpal tunnel syndrome diagnosis. This overcomes manual tracking limitations, enhancing clinical utility.
Area of Science:
- Medical imaging
- Deep learning applications
- Neurology
Background:
- Dynamic sonography aids entrapment neuropathy diagnosis by visualizing nerve movement.
- Manual nerve tracking in ultrasound is labor-intensive and limits clinical adoption.
Purpose of the Study:
- To evaluate automated median nerve segmentation in dynamic sonography using deep learning.
- To assess the performance of various deep learning models for this task.
Main Methods:
- Deep learning models (DeepLabV3+, U-Net, FPN, Mask-R-CNN) were trained and tested on dynamic ultrasound videos of the median nerve from 52 carpal tunnel syndrome patients.
- Performance was quantified using Intersection over Union (IoU) scores.
Main Results:
- DeepLabV3+ and Mask-R-CNN achieved the highest performance, with average IoU scores near 0.83.
- Automated segmentation accurately determined nerve centroid, circularity, perimeter, and cross-sectional area.
Conclusions:
- Deep learning enables feasible and accurate automated median nerve segmentation in dynamic sonography.
- This technology has the potential to improve the clinical diagnosis of carpal tunnel syndrome.
