Related Experiment Video
Updated: Oct 13, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.0K
Group-wise context selection network for choroid segmentation in optical coherence tomography.
Fei Shi1, Xuena Cheng1, Shuanglang Feng1
1MIPAV Lab, the School of Electronics and Information Engineering, Soochow University, Suzhou 215006, People's Republic of China.
Physics in Medicine and Biology
|November 17, 2021
Summary
A new network, GCS-Net, accurately segments the choroid in optical coherence tomography (OCT) images for retinal disease management. It achieves high precision in segmenting choroid thickness, crucial for conditions like high myopia.
Area of Science:
- Ophthalmology
- Medical Imaging
- Computer Vision
Background:
- Choroid thickness measurement from optical coherence tomography (OCT) is essential for managing retinal diseases, particularly high myopia.
- Accurate choroidal segmentation is challenging due to variable thickness and pathological retinal shapes.
Purpose of the Study:
- To introduce a novel group-wise context selection network (GCS-Net) for accurate choroid segmentation in OCT images.
- To address the challenges of diverse choroid thickness and variable pathological retinal shapes.
Main Methods:
- GCS-Net utilizes group-wise channel dilation (GCD) and group-wise spatial dilation modules for multi-scale information selection guided by attention mechanisms.
- A boundary optimization network with edge loss and deep supervision is incorporated to refine choroid boundary segmentation.
Main Results:
- GCS-Net achieved a Dice similarity coefficient of 95.97 ± 0.54% on a dataset of 1650 OCT B-scans.
- The proposed method demonstrated superior performance compared to existing state-of-the-art segmentation networks.
Conclusions:
- GCS-Net offers a robust and accurate solution for choroid segmentation in OCT images.
- The network's design effectively handles variations in choroid thickness and retinal pathology, improving diagnostic capabilities for retinal diseases.
