Related Experiment Video
Updated: Jul 3, 2026
09:17
Using Retinal Imaging to Study Dementia
Published on: November 6, 2017
21.8K
Non-invasive technique to detect diabetic retinopathy based on Electrooculography signal using machine learning
R Archana1, T Rajalakshmi2, P Vijay Sai1
1Department of Biomedical Engineering, SRMIST, Kattankulathur, Tamil Nadu, India.
Summary
A novel, low-cost Electrooculogram (EOG) device non-invasively detects diabetic retinopathy. This method uses statistical features and Support Vector Machine classification, achieving high accuracy for early diagnosis and blindness prevention.
Area of Science:
- Ophthalmology
- Biomedical Engineering
- Medical Diagnostics
Background:
- Diabetic retinopathy is a leading cause of blindness in adults over 40 with diabetes.
- Early detection is crucial for preventing vision loss.
- Current diagnostic methods may be invasive or costly.
Purpose of the Study:
- To develop a low-cost, miniaturized hardware circuit for Electrooculogram (EOG) signal acquisition.
- To classify EOG signals for the early detection of diabetic retinopathy.
- To validate a non-invasive diagnostic tool for diabetic retinopathy.
Main Methods:
- Designed a single-channel EOG hardware circuit with second-order filters.
- Extracted statistical features (kurtosis, mean, std dev, etc.) from software-filtered EOG signals.
- Employed Support Vector Machine (SVM) for signal classification.
Main Results:
- Achieved 93.33% accuracy in classifying normal and diabetic retinopathy subjects.
- SVM classifier demonstrated high sensitivity (96.43%) and specificity (90.625%).
- Area Under Curve (AUC) reached 0.93, indicating favorable diagnostic performance.
Conclusions:
- The proposed EOG-based method is a cost-effective and non-invasive approach for diabetic retinopathy diagnosis.
- The developed prototype and processing methodology show significant potential for clinical application.
- This technology offers an affordable diagnostic tool for diabetic patients at risk of retinopathy.
