Related Experiment Video
Updated: Jan 9, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.3K
STB-Net: a Siamese architecture-based reconstruction-segmentation network for ocular surface image segmentation.
Cheng Wan1,2, Jimei Wu1, Yulong Mao1
1College of Electronic and Information Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing, China.
Quantitative Imaging in Medicine and Surgery
|December 10, 2025
Summary
This study introduces STB-Net, an automated model for precise eyelid measurements from ocular images. The novel deep learning approach significantly improves the diagnosis of eyelid disorders.
Area of Science:
- Ophthalmology
- Medical Imaging
- Computer Vision
Background:
- Accurate eyelid morphological parameters are vital for diagnosing eyelid disorders.
- Automated and precise measurement of these parameters remains a significant clinical challenge.
Purpose of the Study:
- To develop an automated segmentation model for ocular surface images.
- To accurately segment key anatomical structures for precise eyelid metric computation.
Main Methods:
- Proposed STB-Net, a novel segmentation model enhancing TransUNet with a Bottom-up Local Attention Modulation (BLAM) module.
- Integrated TB-Net with SRSNetwork for augmented reconstruction-task training, improving segmentation.
- Model automatically computes palpebral fissure heights, width, and area.
Main Results:
- Achieved high efficacy on a local dataset for palpebral fissure segmentation (Dice: 0.9875, GA: 0.9955, IoU: 0.9767).
- Demonstrated strong corneal segmentation performance (Dice: 0.9891, GA: 0.9978, IoU: 0.9790).
Conclusions:
- STB-Net offers a robust solution for automated ocular surface segmentation.
- Enables precise quantification of eyelid morphological parameters, enhancing diagnostic objectivity and efficiency for eyelid disorders.
