Abstract
Purpose :
Ocular redness is an essential biomarker for many ocular surface diseases for which clinical grading is still done in person by experts using different grading scales, e.g., Efron. This work aims to improve this subjective and time-consuming workflow by calculating image-based redness scores for images acquired using a novel prototype for standardized ocular surface photography. In addition, the scores are compared to clinical gradings.
Methods :
Images of the eyes of healthy subjects (N=18) and subjects with various ocular surface diseases (N=21) were recorded using the novel imaging device. While experts graded the redness of the pathological eyes from normal (0) to severe (4) using the Efron scale, the redness of the healthy cases was assumed to be all normal. Fixed regions-of-interest were defined for all images based on the iris center and split into non-overlapping, squared sub-regions. From these tiles (N=4560), those were selected by two observers where only the ocular surface was visible, see Fig. 1. Finally, intensity-based redness scores in the range of -0.5 and 1 were calculated for each ocular surface classified tile and compared to the clinical gradings.
Results :
The redness scores for the healthy subjects showed lower values and variation than the pathological ones. The mean intensity-based redness was measured to be 0.0255 ± 0.0094 and 0.0598 ± 0.0201 for the healthy and the patient group, respectively. Exemplary images of eyes that were graded between normal and severe, along with the evaluated region-of-interest and scores, are shown in Fig. 1 A to C, respectively. A positive trend was determined when evaluating the objective image-based redness scores against the subjective Efron grades.
Conclusions :
Image-based redness measurement for standardized ocular surface images improves objective ocular surface redness grading. As a result, it may impact clinical time, patient flow, repeatability and reproducibility.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.