Diabetic Retinopathy is a Diabetes complication and the leading cause of blindness in adults in the world. If detected early it can be treated by laser surgery, however its early detection is frequently missed since it progresses without symptoms until irreversible vision loss occurs. Poor glycemic control and hypertension may be manifested by changes at the retinal microvascularity, damaging the blood vessels of the light-sensitive tissue at the back of the eye (retina). These changes lead to a higher degree of vessel permeability, which in turn causes fluid leaks. These leakages, depending on the exact location where they occur can degrade visual acuity. The continuous monitoring is therefore essential to protect the vision, but Retinal Photography (Fundoscopy) is expensive and not easily available in developing countries, and not even in some developed countries.
This project follows the successfully achieved results of the EyeFundusScope, focused on the first detectable signs of the pathology called microaneurysms, detected by image processing algorithms running in smartphone with a low-cost ophthalmoscope mounted to the built-in camera. Current work resulting from EyeFundusScope already includes: a mobile app to capture images from the retina, optic disk tracking, vascular segmentation, microaneurysm detector. However, the detection methods needs to be completed with more features and a decision-support system.
Exudates are often formed as a result of slightly more advanced stages, and their detection is used for a robust evaluation of the patient condition. In particular, soft exudates, also called cotton wool spots, are relevant findings which are still part of nonproliferative (treatable) retinopathy. The main objective of this project is to find these cotton wool spots by texture-based feature extraction and to build a decision-support system for risk assessment of Diabetic Retinopathy. Currently, there is no similar solution to be used by non-specialist practitioner’s end-users, for automatic annotation of retinopathy by smartphones.
This project aims at developing a smartphone-based low computational-cost algorithms and yet highly efficient in the lower quality images of the smartphone camera, improved by multi-image denoising techniques. In addition, the decision-support system can be extended to other eye diseases and stand as a useful tool for eye health screening in developing countries.
> Image recognition and automatic annotation of cotton wool spots on eye fundus in smartphone fundography;
> Decision-support systems for risk assessment of Diabetic Retinopathy based on microaneurysm and cotton wool spots;
> All image processing algorithms must be running in Android platform and adapted to efficiently use the processing power of current smartphones.
These outcomes can be applied in both AAL (Chronic Diseases and Well-being Management) and ICT4D context.
For any additional information regarding this project, please contact us using the inquiries form.