Hi, my name is Parisa Emami. I'm an associate professor of ophthalmology at the University of California, Davis in Sacramento. Here at ARVO 2026 in Denver, I presented our data on OCT-based evaluation of retinal inflammation in patients with uveitis. Uveitis is a common blinding condition. In our current practice, fluorescein angiography or FA is the gold standard for evaluation of retinal inflammation and retinal vasculitis in uveitis patients. However, it is invasive, it is time-consuming and impractical for frequent monitoring of patients in clinic. It is also not available in remote areas and outside of designated and dedicated retinal and uveitis clinics. So we wanted to see if OCT, which is commonly available, denotes the same codes and the same signals for evaluation of uveitis and inflammation in these patients. So we trained a machine learning algorithm using data from non-infectious uveitis patients who underwent same day fluorescein angiography and OCT.

We trained a machine learning model using this data and tested this model on an internal tested from UC Davis and an external test set to evaluate the cross-site generalizability. We found that our model had a great performance with an area under the receiving operator curve of 0.98 in classifying whether the OCTB scan is passing through an inflamed area or it's not passing through an inflamed area, basically a binary classification. After this binary classification, we wanted to translate these findings into a continuous severity index, which we call Odyssey or ODYSSEY, OCT Derived Inflammation Severity Index. This is basically an index showing a number for inflammation over a specific region of the area. Basically, we are quantifying inflammation over a specific region of the retina, which can be used for longitudinal monitoring of these patients. And we found that our ODYSSEY score had a very high correlation with expert derived annotation and expert direct severity index.

So in conclusion, we found that our machine learning algorithm has the potential to diagnose retinal inflammation and retinal vasculitis by just looking at OCT. We were also able to find a continuous inflammation score for longitudinal patient care and for future use as an endpoint in clinical trials. So we are planning on performing future multicenter multi-platform studies to evaluate the efficacy of this platform and these algorithms for using clinic and future use in clinical trials. Thank you.