Related Experiment Video
Updated: Jul 6, 2026
14:08
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
Published on: April 13, 2013
42.6K
Automatic Detection and Classification of Hypertensive Retinopathy with Improved Convolution Neural Network and
Usharani Bhimavarapu1, Nalini Chintalapudi2, Gopi Battineni2
1Department of Computer Science and Engineering, Koneru Lakshmaiah Education Foundation, Vaddeswaram 522302, India.
Bioengineering (Basel, Switzerland)
|January 22, 2024
Summary
This study introduces an automated system for detecting hypertensive retinopathy (HR) severity using retinal images. The novel approach significantly reduces processing time and achieves high accuracy in classifying HR stages.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Hypertensive retinopathy (HR) is a leading cause of preventable blindness worldwide, stemming from hypertension-induced microvascular retinal changes.
- Automated detection and grading of HR severity from retinal images are crucial for timely intervention and vision preservation.
Purpose of the Study:
- To develop and evaluate an automated system for identifying and categorizing the severity of hypertensive retinopathy using retinal fundus images.
- To improve computational efficiency in HR detection compared to existing models.
Main Methods:
- A novel spatial convolution module (SCM) network was developed, integrating cross-channel and spatial information for feature extraction.
- The model was trained and validated on publicly available datasets (ODIR, INSPIREVR, VICAVR) with data augmentation applied to 1200 fundus images.
- Classification of HR severity levels (normal, mild, moderate, severe, malignant) was performed using improved Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) classifiers.
Main Results:
- The proposed model demonstrated reduced processing time due to convolutional layers running only once.
- The improved SVM classifier achieved the highest accuracy of 98.99% in vessel classification, completing the task in 160.4 seconds.
- Ten-fold cross-validation yielded the highest accuracy of 98.99%, outperforming five-fold classification.
Conclusions:
- The developed automated system offers a computationally efficient and accurate method for detecting and classifying hypertensive retinopathy severity.
- The SCM-based approach shows significant potential for clinical application in managing hypertension-related eye disease.
- High accuracy rates achieved by improved SVM and KNN classifiers underscore the effectiveness of the proposed methodology.
