Related Experiment Video
Updated: May 31, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.6K
A Novel Ensemble Meta-Model for Enhanced Retinal Blood Vessel Segmentation Using Deep Learning Architectures
Mohamed Chetoui1, Moulay A Akhloufi1
1Perception, Robotics, and Intelligent Machines Lab (PRIME), Department of Computer Science, Université de Moncton, Moncton, NB E1A 3E9, Canada.
Biomedicines
|January 25, 2025
Summary
This study introduces an ensemble deep learning framework for accurate retinal blood vessel segmentation, improving diagnosis of eye diseases. The novel approach enhances detection of fine vessels, even in noisy images.
Area of Science:
- Medical Imaging
- Computer Vision
- Artificial Intelligence
Background:
- Retinal blood vessel segmentation is crucial for diagnosing diabetic retinopathy, glaucoma, and hypertensive retinopathy.
- Challenges include image noise, low contrast, and complex vessel structures.
Purpose of the Study:
- To develop a robust ensemble learning framework for enhanced retinal blood vessel segmentation.
- To improve the accuracy and generalization of segmentation models.
Main Methods:
- A novel ensemble framework combining U-Net, ResNet50, U-Net with ResNet50 backbone, and U-Net with a transformer block.
- Models were trained on DRIVE and STARE datasets.
- A stacking approach was used to integrate predictions from individual architectures.
Main Results:
- The ensemble model achieved state-of-the-art performance with 0.9778 accuracy, 0.9912 AUC, and 0.8231 F1-Score.
- Demonstrated superior performance in identifying thin retinal blood vessels.
Conclusions:
- The proposed ensemble framework is robust and effective for retinal blood vessel segmentation.
- Outperforms individual models, particularly in challenging conditions like noise and poor visibility.
