Related Experiment Video
Updated: May 5, 2026
07:11
Assessing Early Stage Open-Angle Glaucoma in Patients by Isolated-Check Visual Evoked Potential
Published on: May 25, 2020
7.7K
VisionGuard: enhancing diabetic retinopathy detection with hybrid deep learning
Sudesh Rao1, Sudesh Rao2, Sanjeev D Kulkarni1
1Institute of Engineering and Technology, Srinivas University, Mangalore, India.
Expert Review of Medical Devices
|March 29, 2025
Summary
A new deep learning model, MobileFusionNet, accurately detects diabetic retinopathy (DR) using mobile devices. This automated system achieves 98.19% accuracy, improving early DR detection and preventing vision loss.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Diabetic retinopathy (DR) is a leading cause of preventable blindness.
- Early detection and intervention are crucial for preserving vision.
- Deep learning offers promising avenues for automated DR screening.
Purpose of the Study:
- To develop MobileFusionNet, a novel deep learning model for automated DR detection.
- To integrate MobileNet and GoogleNet architectures for enhanced DR detection.
- To enable DR screening using mobile devices.
Main Methods:
- Implementation in Python using large-scale annotated retinal image datasets.
- Image pre-processing followed by Histogram of Oriented Gradients (HOG) feature extraction.
- Dimensionality reduction using Linear Discriminant Analysis (LDA).
Main Results:
- The MobileFusionNet model demonstrates low inference time and high energy efficiency.
- Achieved high sensitivity and specificity in DR detection.
- Attained an overall accuracy of 98.19% for DR detection.
Conclusions:
- MobileFusionNet offers a modular and easily integrable solution for DR detection.
- The model has significant potential to democratize access to timely and accurate DR screening.
- Particularly beneficial for resource-limited settings to prevent vision loss.
