Related Experiment Video
Updated: Aug 31, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.9K
Atrous residual convolutional neural network based on U-Net for retinal vessel segmentation
Jin Wu1, Yong Liu1,2, Yuanpei Zhu3
1School of Information Science and Engineering, Wuhan University of Science and Technology, Wuhan, China.
Plos One
|August 22, 2022
Summary
This study introduces an improved deep learning model for retinal vessel segmentation. The novel ARN model with an atrous block enhances diagnostic accuracy for eye diseases.
Area of Science:
- Ophthalmology
- Medical Imaging
- Computer-Aided Diagnosis
Background:
- Retinal vessel segmentation is crucial for diagnosing diseases like diabetes and hypertension.
- Current deep learning methods struggle with accuracy due to limited data and complex vessel structures.
Purpose of the Study:
- To propose an effective deep learning model for accurate retinal vessel segmentation.
- To address challenges in segmenting small blood vessels and optic disk areas.
Main Methods:
- Developed an Attention Residual Network (ARN) model incorporating an atrous block.
- Utilized residual convolution networks to increase model depth and performance.
- Evaluated model feasibility using sensitivity, specificity, F1-score, accuracy, and AUC.
Main Results:
- Achieved high accuracy rates of 0.9686 on the DRIVE dataset and 0.9746 on the CHASE DB1 dataset.
- Demonstrated the effectiveness of the proposed ARN model on benchmark datasets.
- The segmentation structure aids physicians in more effective diagnosis.
Conclusions:
- The proposed ARN model with an atrous block significantly improves retinal vessel segmentation accuracy.
- The model's ability to handle complex structures enhances its clinical utility.
- This advancement supports more effective computer-aided diagnosis of eye-related diseases.
