Related Experiment Video
Updated: Sep 24, 2025
Author Spotlight: An Automated Method for Assessing Visual Acuity in Infants and Toddlers Using an Eye-Tracking System
Published on: March 17, 2023
Automatic zoning for retinopathy of prematurity with semi-supervised feature calibration adversarial learning
Yuanyuan Peng1, Zhongyue Chen1, Weifang Zhu1
1MIPAV Lab, School of Electronics and Information Engineering, Soochow University, Suzhou, Jiangsu 215006, China.
Insights
This study introduces a new semi-supervised learning network for retinopathy of prematurity (ROP) zoning, improving diagnosis accuracy for this leading cause of childhood blindness. The novel approach effectively uses unlabeled data to enhance ROP classification.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Retinopathy of prematurity (ROP) is a significant cause of childhood blindness in premature infants.
- Current automated ROP diagnosis primarily focuses on screening and disease presence, with limited research on ROP zoning.
- High-quality annotations for ROP zoning are scarce, hindering model development and effective utilization of unlabeled data.
Purpose of the Study:
- To develop a novel semi-supervised feature calibration adversarial learning network (SSFC-ALN) for accurate 3-level ROP zoning.
- To address the challenge of limited labeled data in ROP diagnosis by leveraging unlabeled data through semi-supervised learning.
- To improve the classification performance of ROP zoning by incorporating an attention-based feature calibration module.
Main Methods:
- Proposed a semi-supervised feature calibration adversarial learning network (SSFC-ALN) comprising a generative U-net and a compound network with a discriminator.
- Integrated an attention mechanism-based feature calibration module (FCM) into the compound network to enhance focus on relevant image features.
- Employed a 3-fold cross-validation strategy on a dataset of 1013 fundus images from 108 patients.
Main Results:
- The SSFC-ALN achieved high classification performance for 3-level ROP zoning.
- The proposed method demonstrated superior results compared to other state-of-the-art classification techniques.
- The integration of the FCM and adversarial learning effectively improved the diagnostic accuracy.
Conclusions:
- The developed SSFC-ALN is a promising approach for automated ROP zoning, particularly in scenarios with limited labeled data.
- The study highlights the effectiveness of semi-supervised learning and attention mechanisms in improving ROP diagnosis.
- This work contributes to advancing objective and accurate assessment of ROP severity through automated fundus image analysis.
Abstract:
Retinopathy of prematurity (ROP) is an eye disease, which affects prematurely born infants with low birth weight and is one of the main causes of children's blindness globally. In recent years, there are many studies on automatic ROP diagnosis, mainly focusing on ROP screening such as "Yes/No ROP" or "Mild/Severe ROP" and presence/absence detection of "plus disease". Due to the lack of corresponding high-quality annotations, there are few studies on ROP zoning, which is one of the important indicators to evaluate the severity of ROP. Moreover, how to effectively utilize the unlabeled data to train model is also worth studying. Therefore, we propose a novel semi-supervised feature calibration adversarial learning network (SSFC-ALN) for 3-level ROP zoning, which consists of two subnetworks: a generative network and a compound network. The generative network is a U-shape network for producing the reconstructed images and its output is taken as one of the inputs of the compound network. The compound network is obtained by extending a common classification network with a discriminator, introducing adversarial mechanism into the whole training process. Because the definition of ROP tells us where and what to focus on in the fundus images, which is similar to the attention mechanism. Therefore, to further improve classification performance, a new attention mechanism based feature calibration module (FCM) is designed and embedded in the compound network. The proposed method was evaluated on 1013 fundus images of 108 patients with 3-fold cross validation strategy. Compared with other state-of-the-art classification methods, the proposed method achieves high classification performance.
