Related Experiment Video
Updated: Aug 2, 2025
11:29
Corneal Confocal Microscopy: A Novel Non-invasive Technique to Quantify Small Fibre Pathology in Peripheral Neuropathies
Published on: January 3, 2011
26.7K
Multi-scale and local feature guidance network for corneal nerve fiber segmentation
Wei Tang1, Xinjian Chen1,2, Jin Yuan3
1MIPAV Lab, School of Electronic and Information Engineering, Soochow University, People's Republic of China.
Physics in Medicine and Biology
|April 13, 2023
Summary
This study introduces MLFGNet, a novel neural network for automatically segmenting corneal nerve fibers in eye images. The method shows excellent performance, aiding in early diagnosis of neurological diseases like diabetic neuropathy.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Corneal confocal microscopy (CCM) is crucial for visualizing corneal nerves.
- Accurate segmentation of corneal nerve fibers is essential for diagnosing systemic neurological diseases, including diabetic peripheral neuropathy.
- Current segmentation methods require improvement for clinical application.
Purpose of the Study:
- To develop an automated method for segmenting corneal nerve fibers in CCM images.
- To enhance the accuracy and efficiency of corneal nerve fiber analysis.
- To improve early diagnosis of neurological disorders through advanced imaging analysis.
Main Methods:
- A U-shape encoder-decoder neural network, MLFGNet, was designed for corneal nerve fiber segmentation.
- Novel modules (MFPG, LFGA, MDS) were integrated to improve multi-scale and local feature extraction.
- The network focuses on fusing multi-scale information and extracting local features for enhanced discrimination of nerve fiber structures.
Main Results:
- MLFGNet achieved high Dice coefficients of 89.33%, 89.41%, and 88.29% on three CCM image datasets.
- The proposed method demonstrated superior segmentation performance compared to existing state-of-the-art techniques.
- The novel modules effectively addressed information imbalances and improved feature reconstruction.
Conclusions:
- MLFGNet offers excellent performance for corneal nerve fiber segmentation in CCM images.
- The developed method has the potential to significantly aid in the early detection of neurological diseases.
- This automated approach advances the clinical utility of CCM for diagnosing systemic conditions.
