Related Experiment Video
Updated: Dec 11, 2025
11:29
Corneal Confocal Microscopy: A Novel Non-invasive Technique to Quantify Small Fibre Pathology in Peripheral Neuropathies
Published on: January 3, 2011
27.1K
A Deep Learning Model for Automated Sub-Basal Corneal Nerve Segmentation and Evaluation Using In Vivo Confocal
Shanshan Wei1,2, Faqiang Shi3,4,5, Yuexin Wang1,2
1Department of Ophthalmology, Peking University Third Hospital, Beijing, China.
Translational Vision Science & Technology
|August 25, 2020
Summary
A new deep learning model, CNS-Net, accurately segments and evaluates corneal nerve fibers (CNF) from in vivo confocal microscopy (IVCM) images. This automated system offers a faster alternative to manual analysis for clinical applications.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Corneal nerve fiber (CNF) analysis is crucial for diagnosing ocular surface diseases.
- Manual segmentation and evaluation of CNF from in vivo confocal microscopy (IVCM) images are time-consuming and subjective.
Purpose of the Study:
- To develop and validate a deep learning model for automated segmentation and evaluation of sub-basal CNF using IVCM images.
Main Methods:
- A convolutional neural network-based model, CNS-Net, was developed for CNF segmentation.
- The model was trained on 552 labeled IVCM images and tested on 139 images.
- Performance was evaluated using AUC, mAP, sensitivity, specificity, and relative deviation ratio (RDR).
Main Results:
- CNS-Net achieved high accuracy with an AUC of 0.96 and mAP of 94% for segmentation.
- The model demonstrated 96% sensitivity and 75% specificity in CNF segmentation.
- CNFL evaluation had an RDR of 16%, and the model processed images at 32 frames per second.
Conclusions:
- The developed CNS-Net model provides accurate and rapid automated segmentation and evaluation of sub-basal CNF from IVCM images.
- This deep learning approach shows significant potential to aid clinical diagnosis and treatment of ocular surface diseases.
