Related Experiment Video
Updated: Jul 30, 2025
09:17
Using Retinal Imaging to Study Dementia
Published on: November 6, 2017
21.7K
Mobile-HR: An Ophthalmologic-Based Classification System for Diagnosis of Hypertensive Retinopathy Using Optimized
Muhammad Zaheer Sajid1, Imran Qureshi2, Qaisar Abbas2
1Department of Computer Software Engineering, MCS, National University of Science and Technology, Islamabad 44000, Pakistan.
Diagnostics (Basel, Switzerland)
|May 16, 2023
Summary
A new Mobile-HR system accurately diagnoses hypertensive retinopathy (HR) using transfer learning and dense blocks. This lightweight approach achieves 99% accuracy, aiding early detection and preventing vision loss from high blood pressure effects on the eyes.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Hypertensive retinopathy (HR) is a serious eye condition caused by high blood pressure, leading to changes in retinal arteries.
- Early detection of HR is crucial to prevent vision loss, with diagnosis often relying on ophthalmologists analyzing fundus images.
- Existing computer-aided diagnosis (CADx) systems using deep learning face challenges like large dataset requirements, class imbalance, and overfitting.
Purpose of the Study:
- To develop a lightweight and efficient CADx system for diagnosing hypertensive retinopathy.
- To address limitations of current deep learning models, such as computational complexity and the need for extensive datasets.
- To improve the accuracy and performance of automated HR detection.
Main Methods:
- Developed a novel Mobile-HR system integrating a pretrained transfer learning (TL) MobileNet architecture with dense blocks.
- Employed data augmentation techniques to increase the size and diversity of training and testing datasets.
- Optimized the network for diagnosing HR eye disease, focusing on lightweight feature descriptors and computational efficiency.
Main Results:
- The Mobile-HR system achieved a high accuracy of 99% and an F1 score of 0.99 across various datasets.
- The proposed approach demonstrated superior performance compared to existing state-of-the-art HR detection methods.
- Experimental results were validated by an expert ophthalmologist, confirming the system's effectiveness.
Conclusions:
- The developed Mobile-HR system offers a highly accurate and efficient solution for diagnosing hypertensive retinopathy.
- This lightweight CADx model effectively overcomes challenges associated with small datasets and computational complexity.
- The findings suggest significant potential for Mobile-HR in clinical settings for early and accurate HR detection, aiding in vision preservation.
Keywords:
DenseNetMobileNetcomputer-aided diagnosisconvolutional neural networkdeep learninghypertensive retinopathyretinal fundus imagestransfer learning
