Related Experiment Video
Updated: Nov 13, 2025
Monitoring Dynamic Growth of Retinal Vessels in Oxygen-Induced Retinopathy Mouse Model
Published on: April 2, 2021
Automatic Staging for Retinopathy of Prematurity With Deep Feature Fusion and Ordinal Classification Strategy
Insights
This study introduces a novel deep neural network for accurate 5-level staging of retinopathy of prematurity (ROP) in premature infants. The method effectively classifies ROP stages from retinal images, aiding in preventing childhood blindness.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Retinopathy of prematurity (ROP) is a leading cause of preventable childhood blindness in premature infants.
- Current research often focuses on ROP screening and plus disease detection, with limited attention to precise ROP staging.
- Accurate ROP staging is crucial for effective disease management and severity evaluation.
Purpose of the Study:
- To develop and validate a novel deep neural network for precise 5-level staging of retinopathy of prematurity (ROP).
- To improve the accuracy and clinical applicability of automated ROP staging using fundus images.
Main Methods:
- A multi-stream parallel feature extractor utilizing ResNet18, DenseNet121, and EfficientNetB2 to capture diverse high-level features.
- Deep feature fusion through concatenation and convolution to create a comprehensive feature representation.
- An ordinal classification strategy tailored for clinical 5-level ROP staging.
Main Results:
- The proposed ROP staging network achieved high performance metrics, including weighted recall (0.9055), precision (0.9092), and F1 score (0.9043) for per-image staging.
- Excellent accuracy (ACC1: 0.9827) and Kappa (0.9786) were obtained for per-image ROP staging on a dataset of 635 images.
- Validation on 1173 examinations using 4-fold cross-validation demonstrated the method's validity and advantages for per-examination ROP staging.
Conclusions:
- The proposed deep neural network effectively performs 5-level ROP staging, addressing a gap in current automated diagnostic tools.
- The multi-stream feature extraction, fusion, and ordinal classification approach enhance ROP staging accuracy.
- This method shows significant potential for clinical application in managing retinopathy of prematurity and preventing childhood blindness.
Abstract:
Retinopathy of prematurity (ROP) is a retinal disease which frequently occurs in premature babies with low birth weight and is considered as one of the major preventable causes of childhood blindness. Although automatic and semi-automatic diagnoses of ROP based on fundus image have been researched, most of the previous studies focused on plus disease detection and ROP screening. There are few studies focusing on ROP staging, which is important for the severity evaluation of the disease. To be consistent with clinical 5-level ROP staging, a novel and effective deep neural network based 5-level ROP staging network is proposed, which consists of multi-stream based parallel feature extractor, concatenation based deep feature fuser and clinical practice based ordinal classifier. First, the three-stream parallel framework including ResNet18, DenseNet121 and EfficientNetB2 is proposed as the feature extractor, which can extract rich and diverse high-level features. Second, the features from three streams are deeply fused by concatenation and convolution to generate a more effective and comprehensive feature. Finally, in the classification stage, an ordinal classification strategy is adopted, which can effectively improve the ROP staging performance. The proposed ROP staging network was evaluated with per-image and per-examination strategies. For per-image ROP staging, the proposed method was evaluated on 635 retinal fundus images from 196 examinations, including 303 Normal, 26 Stage 1, 127 Stage 2, 106 Stage 3, 61 Stage 4 and 12 Stage 5, which achieves 0.9055 for weighted recall, 0.9092 for weighted precision, 0.9043 for weighted F1 score, 0.9827 for accuracy with 1 (ACC1) and 0.9786 for Kappa, respectively. While for per-examination ROP staging, 1173 examinations with a 4-fold cross validation strategy were used to evaluate the effectiveness of the proposed method, which prove the validity and advantage of the proposed method.
