Related Experiment Video
Updated: Oct 29, 2025
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
Assistive Framework for Automatic Detection of All the Zones in Retinopathy of Prematurity Using Deep Learning
Ranjana Agrawal1,2, Sucheta Kulkarni3, Rahee Walambe4
1School of Computer Engineering and Technology, Dr. Vishwanath Karad MIT World Peace University, Pune, India.
Insights
A new automated system accurately identifies retinal zones in premature infants, aiding in timely diagnosis and treatment of retinopathy of prematurity (ROP). This technology helps prevent blindness by improving ROP screening and follow-up scheduling.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Retinopathy of prematurity (ROP) poses a significant risk of blindness in preterm infants, particularly in India.
- Current automated screening systems lack the ability to identify specific retinal zones crucial for ROP management.
- Accurate ROP staging requires identification of vascularized retinal zones (Zones I, II, III) for appropriate follow-up and discharge decisions.
Purpose of the Study:
- To develop a novel automated system for detecting retinal Zones I, II, and III in infants with underdeveloped macula.
- To address challenges in ROP image analysis, including low contrast, incomplete vascularization, and lack of public datasets.
- To assist medical experts in interpreting retinal vascular status and reducing diagnostic subjectivity.
Main Methods:
- Implementation of a novel method utilizing an ensemble of U-Network and Circle Hough Transform.
- Training the generic model on diverse retinal images of variable sizes and qualities.
- Evaluation on a test set including low-quality images captured by different imaging systems.
Main Results:
- The developed system achieved 98% accuracy in detecting retinal zones (I, II, III) across variable image sizes and qualities.
- The system demonstrated robustness, performing accurately on all test images, including low-quality ones.
- Rapid processing times: 14 minutes for training and 30 milliseconds for single image testing.
Conclusions:
- The novel ensemble method effectively detects retinal zones crucial for ROP staging and management.
- This automated system can significantly aid clinicians in accurate ROP diagnosis and reduce inter-observer variability.
- The technology holds potential for improving ROP screening efficiency and preventing visual impairment in at-risk infants.
Abstract:
Retinopathy of prematurity (ROP) is a potentially blinding disorder seen in low birth weight preterm infants. In India, the burden of ROP is high, with nearly 200,000 premature infants at risk. Early detection through screening and treatment can prevent this blindness. The automatic screening systems developed so far can detect "severe ROP" or "plus disease," but this information does not help schedule follow-up. Identifying vascularized retinal zones and detecting the ROP stage is essential for follow-up or discharge from screening. There is no automatic system to assist these crucial decisions to the best of the authors' knowledge. The low contrast of images, incompletely developed vessels, macular structure, and lack of public data sets are a few challenges in creating such a system. In this paper, a novel method using an ensemble of "U-Network" and "Circle Hough Transform" is developed to detect zones I, II, and III from retinal images in which macula is not developed. The model developed is generic and trained on mixed images of different sizes. It detects zones in images of variable sizes captured by two different imaging systems with an accuracy of 98%. All images of the test set (including the low-quality images) are considered. The time taken for training was only 14 min, and a single image was tested in 30 ms. The present study can help medical experts interpret retinal vascular status correctly and reduce subjective variation in diagnosis.
