SUMMARY:
UCLA researchers in the Department of Ophthalmology have developed a deep learning-based automated algorithm that can grade funduscopic images for the diagnosis and treatment of glaucoma.
BACKGROUND:
Glaucoma is the second leading cause of blindness in the world. In 2020, about 80 million patients suffered from glaucoma, and over 110 million are projected to have the disease by 2040. To ensure proper treatment of the disease, diagnosis and progression monitoring are critical, and can be accomplished via several modalities. Retina Tomography (HRT) and Optical Coherence Tomography (OCT) are viable methods, but they are new and often prohibitively expensive for clinics in lower socioeconomic regions. Fundoscopic photos of the optical disc and cup, on the other hand, are commonly used and are widely accessible. These photos are graded to determine the severity and progression of the glaucoma using the Disk Damage Likelihood Scale (DDLS). However, DDLS is implemented by clinicians, and the degree of diagnosis can therefore vary greatly from doctor-to-doctor. Therefore, there is a need for a standard method to perform DDLS scoring for the diagnosis and monitoring of glaucoma.
INNOVATION:
UCLA researchers in the Department of Ophthalmology have developed a standardized method to execute the Disk Damage Likelihood Scale (DDLS) using a deep learning-based system. The method can analyze digital and slide fundoscopic images and output a DDLS score for the diagnosis of glaucoma. The predicted scores were shown to have fair agreement with DDLS grading done by clinicians. Furthermore, the project has collected and will keep updating a large database for improved DDLS scoring. The accumulated data and program could help provide clinics located in lower socioeconomic areas a more accessible way to diagnose glaucoma without the need of expensive equipment.
POTENTIAL APPLICATIONS:
- Age related macular degeneration diagnostics and monitoring
- Cataract diagnosis and monitoring
- Amblyopia diagnosis and monitoring
- Strabismus diagnosis and monitoring
ADVANTAGES:
- Automated scoring
- Deep learning-based pipeline
- Cost effective
- Unique database
DEVELOPMENT-TO-DATE:
First description of complete invention