Abstract
Purpose:
To assess corneal staining induced by exposure to the Controlled Adverse Environment (CAE®) using an automated software-based approach
Methods:
Dry eye subjects (N=195) were exposed for 90 minutes to the CAE®. Clinical assessments and photography of the ocular surface after sodium fluorescein dye instillation were performed before and after challenge. A total of 580 digital images of the inferior cornea were selected and graded automatically. Images were analyzed using software to detect and enumerate stained punctate dots. The clinical score (Ora CalibraTM fluorescein staining scale) was derived from the number of punctate dots as Gpred=1.31log(Nspk) + 0.394, where Nspk was the total dot count for each image. Agreement between clinical and automated methods was assessed statistically with Bland Altman analyses.
Results:
Mean pre-CAE staining scores were 1.99 (SD 0.78) graded by the clinician versus 2.10 (SD 0.85) using the automated analysis; mean post-CAE scores were 2.64 (SD 0.825) graded by the clinician versus 2.52 (SD 0.825) using the automated analysis. The mean inferior corneal staining scores for the entire dataset was 2.32 (SD 0.86) assessed clinically versus 2.31 (SD 0.86) assessed automatically. As evaluated by both methods, the increase in staining with CAE exposure was statistically significant (p<0.003). Bland-Altman analysis showed that the mean score difference between methods was 0.005 (SD 0.726), not significantly different from zero (p>0.7), thereby demonstrating no statistical bias.
Conclusions:
These results indicate that meaningful staining data can be collected in a clinical trial setting using automated software analysis. The software approach provides improved standardization of grading in multi-center clinical studies, and can be used as a supplemental quality assurance tool for the current gold standard of traditional grading at the slit lamp.