SUMMARY
UCLA researchers in the Department of Computer Science have developed a method for the semantic segmentation and quantitative analysis of retinal arteries and veins via infrared reflectance retinal imaging (RAVIR).
BACKGROUND
Retinal vasculature analysis can be used to help diagnose neurodegenerative disorders and monitor systemic diseases such as hypertension and diabetes. Since the retina and its vasculature are optically visible, they can be imaged non-invasively. Current methods for diagnosis and monitoring however, are based on qualitative morphological changes to the vasculature. New automated ways to accurately and reproducibly quantify changes in the eye are needed for improved diagnosis and patient outcomes.
INNOVATION
UCLA researchers in the Department of Computer Science have developed an automated, reliable, and reproducible system for the quantitative assessment of retinal arteries and veins. The system uses IR imaging and a convolutional neural network (CNN) architecture, optimized with a deep learning framework, for the semantic segmentation and assessment of retinal arteries and veins. The method has been successfully prototyped and validated for its effectiveness. The method can accurately segment the retinal arteries and veins from IR images and provide width measurements of extracted vessels in a fully-automated manner. The approach can be expanded to studying morphological changes of retinal arteries to create early predicted models for early stage diagnosis of various diseases.
POTENTIAL APPLICATIONS
- Early disease detection
- neurodegenerative disorders
- Point-of-care
- Patient monitor of disease progression
- Therapeutic monitoring
ADVANTAGES
- Automated
- Reproducible
- Early predictive models
- Non-invasive
DEVELOPMENT TO DATE
Successful demonstration of RAVIR for quantification of retinal vasculature.