The Effect of Photographic Light Exposure on Cup to Disc Ratio Grading


 Background: Digital optic disc images are integral in remote telehealthcare, yet no quality control standards exist for the exposure setting of the images. This study evaluated the relationship between exposure setting and cup/disc ratio (c/d) grading among glaucoma specialists. Methods: Color disc photos were taken of 50 eyes of pediatric patients under anesthesia at 3 light exposure settings: dark, medium, and bright. 9 glaucoma specialists evaluated the c/d of the imaged discs in random order. The relationships between the exposure levels and the c/d estimates as well as between c/d size and variability were evaluated.Results: A total of 150 photos of 50 optic discs were graded. The c/d estimates were significantly larger among bright when compared to photos taken at the medium exposure setting(0.53 vs 0.48, P < 0.001) as well as the dark exposure setting (0.47, P < 0.001). In addition, images with larger mean c/d did not show more variability amongst readers (P = 0.59). Conclusion: Image exposure affects c/d grading of nonstereoscopic disc images. The brighter exposure is associated with larger c/d grading. Exposure consistency is key in longitudinal care.


Background
Digital color fundus photos are integral in remote telehealth ophthalmology care and will play a larger role with the increasing popularity of telehealth. [1] In particular, digital optic disc images will be central to remote glaucoma screening, evaluation, and management. [2] Currently, there are no standardized quality control protocols in obtaining digital optic disc images, with the light exposure setting being one of the variables that can change based on the device and the photographer. The effect of this variability on the intra-and inter-grader assessment of the optic disc con gurations is unknown. [3] In this study, we evaluated the effect of light exposure setting on c/d grading of nonstereoscopic disc images among glaucoma specialists with varying levels of experience.

Methods
Digital optic disc images were obtained using the RetCam (Natus Medical, Inc. Pleasanton, CA, USA) from 50 eyes of 50 pediatric patients during examinations under anesthesia. Prior to the exams, informed consent was provided by all parents or legal guardians to have any images obtained used for research and educational purposes. Each optic disc was imaged at 3 different light exposure settings: dark (RetCam light exposure setting 20), medium (exposure setting 25), and bright (exposure setting 30), with the ambient light of the room kept the same for each capture. Only photo triads with clearly visible thirdorder vessels were kept for the study to ensure su ciently high resolution. The set of 150 photographs were presented to 9 fellowship-trained glaucoma specialist on high-de nition view screens in randomized orders. The 9 glaucoma specialists had practice range between 3 and 42 years (median 6 years). The glaucoma specialists were asked to provide a c/d estimate to the nearest 0.05. A subset of the photographers were repeatedly graded at random orders to assess intraobserver reliability.

Statistics/Data Analysis
Mean c/d gradings for each of the 3 light settings were compared with each of the others using a paired t test. A paired t-test was performed in lieu of both Analysis of Variance (ANOVA) and independent samples t-test because each of the three comparisons is independently relevant, and the same 50 eyes were used for each of the three light exposure groups. A P value of 0.05 or less was considered to be statistically signi cant.

Results
Across all of the photos evaluated by the 9 glaucoma specialists, the mean c/d ratio was 0.50. The mean c/d (+/-SD) for each of the three groups was 0.47 (0.112) in the dark group (n = 50), 0.48 (0.109) in the medium group (n = 50), and 0.53 (0.118) in the bright group (n = 50).
When the 3 photo groups were compared among each other, the c/d readings were signi cantly larger among the bright exposure when compared to the medium exposure photos (0.53 vs 0.48, P < 0.001) as well as the dark exposure (0.47, P < 0.001) ( Table 1). In addition, the mean c/d reading at the medium exposure was larger than that of the dark exposure photos, but this difference failed to reach statistical signi cance (Table 1). between years of training and the variability of the disc readings (P = 0.81) (Fig. 1). In addition, images with larger mean c/d did not show more variability amongst readers (P = 0.59) (Fig. 2).

Discussion
The results of the study showed that the light exposure setting of digital fundus photos impacted glaucoma specialists' grading of the c/d. Among the same patients at the same time points, graders were more likely to assign a higher c/d if the photo was taken at a bright exposure setting. This has implications for both glaucoma screening and the monitoring of an established patient's disease Page 4/6 progression for both patient care and clinical research purposes. While other factors and sophisticated tools play a role in evaluating glaucomatous damage of an optic nerve, the c/d remains one way to quickly estimate the extent of disease in an individual or the prevalence within a population. [4,5] For established patients, inconsistency in the light exposure setting could lead to the over-calling or the under-calling of a patient's glaucoma progression over time or between different providers that may be using different light exposure settings.
These results also suggest that the outcomes of glaucoma screenings may differ based on the light exposure setting of the fundus camera being used, namely, that a brighter exposure setting would increase the mean c/d of the population being screened, increasing the sensitivity but decreasing the speci city of a screening program. For example, in a retrospective cross-sectional study of Haitian Afro-Caribbean patients screened for glaucoma at community health fairs in Miami, FL, patients were classi ed as glaucoma suspects if their c/d exceeded 0.7. [6] Within our sample, 18% of subjects would be classi ed as glaucoma suspects using the dark exposure setting, and 26% would be considered glaucoma suspects using the bright exposure setting. Although the two study populations differ, this illustrates that light exposure levels could impact prevalence estimates and the success of community glaucoma screenings. Since c/d remains a useful measure for glaucoma detection in community screening programs and retrospective studies, a standard protocol for light exposure setting could improve the accuracy of both.

Conclusions
Exposure consistency is key in both care of the individual patient and study of the population. A standardized exposure setting or an algorithm to adjust exposure setting based on a patient's previous photos would ensure that evaluation of optic discs is reliable and valid across individuals and populations. Authors' contributions: ER, AG, and TC were involved in study design and data collection. MM analyzed and interpreted the study data, and wrote the manuscript. TC was a major contributor in the editing of the manuscript. All authors reviewed and approved the nal manuscript. Figure 1 Years of ophthalmology experience versus standard deviation of the cup-to-disc readings. No correlation was observed between the cup-to-disc graders' years of experience and the variability of the cup-to-disc readings that they provided for each of the images. Mean cup-to-disc size versus the standard deviation of the images across all 9 graders No correlation was observed between the mean cup-to-disc size and the variability of the C/D across the 9 graders. Optic discs graded as larger did not show a decreased in the precision of the grading.