Using Artificial Intelligence To Diagnose Uveitis

Summary: 
The National Eye Institute seeks research co-development partners and/or licensees for a deep learning algorithm that can identify retinal vasculitis using color fundus images.

Description of Technology: 
Uveitis is caused by inflammation in the eye that can cause pain and reduce vision. The rate of uveitis in the United States is 1 in every 200 people with eye-related irritation. Permanent symptoms such as vision loss can occur if untreated. Therefore, early detection is crucial. 

In certain uveitis cases, fluorescein angiography (FA) is essential for the diagnosis and management due to its ability to display retinal vascular leakage (RVL). Although proven to be critical in diagnosing and assessing severity, FA is invasive and side effects have been reported. Additionally, the procedure is time-consuming and imposes economic burdens to patients, physicians and payors. 

Scientists at the NEI have developed a deep learning tool to non-invasively detect RVL using ultrawide-field color fundus photos. This algorithm identifies fundus images with and without RVL with high accuracy (79%) and sensitivity (85%). Compared to the current gold standard of assessing RVL (clinician interpretation), this deep learning tool provides an improved method of detecting RVL for patients with uveitis.

Potential Commercial Applications:

• Diagnostic tool to predict uveitis 
• Add-on to current color fundus imaging modalities

Competitive Advantages:

• Greater accuracy and sensitivity versus current gold standard to assess RVL (clinician assessment)
• Deep learning tool to assess RVL
• Deep learning to assess ultrawide-field color fundus images and assess RVL