Related Experiment Video
Updated: May 5, 2026
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
2.3K
A CNN-Transformer fusion network for Diabetic retinopathy image classification
Xuan Huang1, Zhuang Ai2, Chongyang She3
1Medical Research Center, Beijing Chao-Yang Hospital, Capital Medical University, Beijing, China; Department of Ophthalmology, Beijing Chao-Yang Hospital, Capital Medical University, Beijing, China.
Summary
A new AI model, DR-CTFN, accurately detects diabetic retinopathy (DR) from eye images. This advanced deep learning approach offers a faster, more reliable method for diagnosing DR, helping to prevent vision loss.
Area of Science:
- Ophthalmology
- Artificial Intelligence
- Medical Imaging
Background:
- Diabetic retinopathy (DR) is a primary cause of blindness globally.
- Current DR diagnosis relies on manual interpretation of fundus images, which is time-consuming and subjective.
- There is a need for automated, accurate, and scalable DR detection methods.
Purpose of the Study:
- To develop and evaluate a novel deep learning model, DR-CTFN, for automated diabetic retinopathy detection.
- To improve the accuracy and efficiency of DR diagnosis compared to existing methods.
- To provide a robust tool for early detection and prevention of vision loss in DR patients.
Main Methods:
- A convolutional neural network-transformer fusion model (DR-CTFN) was developed, integrating ConvNeXt and Swin Transformer with a lightweight attention block (LAB).
- Standardized preprocessing and extensive image augmentation were employed to address dataset imbalance.
- The model was trained and validated on the Kaggle EyePACS dataset, with external validation on APTOS 2019 and a clinical DR dataset.
Main Results:
- DR-CTFN demonstrated superior performance over standalone ConvNeXt and Swin Transformer models on the Kaggle EyePACS dataset, achieving higher accuracy and Area Under the Curve (AUC).
- External validation showed high accuracy (84.45% and 85.31%) and AUC (95.22% and 95.79%) on independent datasets.
- The model proved effective in rapid, robust, and precise DR detection.
Conclusions:
- The DR-CTFN model offers a significant advancement in automated diabetic retinopathy detection.
- This AI-driven approach provides a scalable solution for early DR diagnosis and prevention of vision impairment.
- The findings suggest DR-CTFN can enhance patient quality of life by enabling timely intervention.
