Thank you for your submission!
Grading Image Scale (G.I.S) for anterior chamber cell count in patients with anterior uveitis.
Neumann Ron, MD, Marc De-Smet, MD, Daphne Haim-Langford, Ph.D, Zohar Milman, MSc, Michal Kremer, MD.
CMO Tarsius Pharma
Purpose: To develop a standardized visual scale to be used by clinicians when assessing anterior chamber cells (ACC). ACC count reliably reflects the severity of both acute and chronic anterior chamber inflammation. While SUN gives a clear categorical definition of ACC, studies have shown that there is significant intra and inter-observer variability which poses a significant challenge in clinical studies. The use of visual scales has been successfully used in trials for the assessment of vitreous haze, diabetic retinopathy, AMD, and cataracts. A scale for ACC seems long overdue.
Methods: A set of computer generated graphical images representing different grades of ACC was created based on the SUN criteria: Grade 0: zero cells, grade 1: 1-5 cells, grade 2: 6-15 cells, grade 3: 16-30 cells and grade 4: more than 30 cells. For each grade, 4 different images where generated.
Uveitis experts were asked to grade the different images and the consistency between graders was calculated. A consistency threshold was set to > 75%. For this purpose, a Tarsius Grading algorithm was implemented using Python in which the user transitions seamlessly between (1) test phase in which the user learns how to score ACC using the visual scale to (2) main scoring phase in which the user grades randomized images representing different ACC grades.
Results: Eleven uveitis experts participated in the evaluation. Their scores were analyzed, and the scoring of the images found to be consistent between different graders. The higher consistency was found for the lower ACC grades of 0, 1 and 2 (98%, 93% and 91% respectively) and lower consistency for ACC grades 3 and 4 (77% and 83% respectively).
Conclusions: Tarsius ACC Grading Image Scale is one step toward an objective, standardized and reliable evaluation of ACC. Further analysis of this tool in a clinical setting is needed.