Related Experiment Video
Updated: Oct 18, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.0K
Multi-path cascaded U-net for vessel segmentation from fundus fluorescein angiography sequential images
Gang Sun1, Xiaoyan Liu2, Xuefei Yu1
1College of Electrical & Information Engineering, Hunan University.
Computer Methods and Programs in Biomedicine
|October 1, 2021
Summary
A novel Multi-Path Cascaded U-net (MCU-net) accurately segments retinal vessels in fundus fluorescein angiography (FFA) images. This method enhances analysis of vascular morphology in various disease stages.
Area of Science:
- Ophthalmology
- Medical Imaging
- Computer Vision
Background:
- Fundus fluorescein angiography (FFA) is crucial for diagnosing retinal diseases.
- Accurate vessel segmentation in FFA images is essential for quantifying vascular morphology.
- Existing segmentation methods struggle with the variable backgrounds and vessel characteristics in sequential FFA images.
Purpose of the Study:
- To develop an advanced vessel segmentation method for FFA sequential images.
- To improve the accuracy and robustness of retinal vessel segmentation in FFA analysis.
Main Methods:
- Proposed a Multi-Path Cascaded U-net (MCU-net) architecture.
- Integrated features from raw FFA images and two synthetic modes enhancing small and large vessels.
- Generated a vascular probability map by fusing multi-modal FFA image features.
Main Results:
- MCU-net achieved superior performance (F1-score, sensitivity, accuracy) compared to state-of-the-art methods on public and private datasets.
- The method effectively preserved fine details like thin vessels and connections.
- Demonstrated robustness across different FFA perfusion stages.
Conclusions:
- The MCU-net provides highly accurate and robust vessel segmentation for FFA sequential images.
- This technique shows significant potential for quantitative vascular morphology analysis in ophthalmology.
