Purpose:
Ocular surface (OSF) vital dye staining is an essential measure of OSF damage. Current grading schemes require real-time observation of dye staining patterns and are hampered by the need to refer to standard diagrams, limited observation time, and inherent observer subjectivity. We studied 2 methods to assess OSF staining via digital imaging: masked human grading and automated assessment.
Methods:
In eyes of patients with ocular surface disease, sodium fluorescein and lissamine green dyes were used via standardized procedures (volume, duration, photographic techniques). Simultaneous slit lamp clinical grading, Oxford scale (range, 0-5) of temporal (T), nasal (N) conjunctiva, and cornea (K) staining and digital photographs (OIS WinStation) were obtained and repeated 1 week later. Subsequent masked reviews of the digital images were done by the original clinical grader and 2 other graders. Digital images were also analyzed using an automated image processing segmentation algorithm consisting of 3 methods to identify area of staining intensity (StInt) for red (R), green (G), and blue (B) channels: 1) RGB: average R StInt/average G StInt; 2) RGB all:ratios (R/G, B/G, and B/R StInt); and 3) levels of HSB (Hue/Saturation/Brightness).
Results:
Data from 107 eyes of 41 subjects (93% female, 31-85 yrs, median 54.3 yrs, 80% White, 10% African American, 10% Asian) were analyzed. Agreement between digital and clinical grades are reported in the Table. StInt scores based on automated image processing increased significantly with increasing clinical grade: RGB, p=0.0001 (T), 0.0009 (N); RGB all, p=0.0001 (T), 0.0016 (N); HSB hue, p=0.0001 (T), 0.003 (N).
Conclusions:
Agreement of masked digital photo grading with gold standard clinical grading was moderate to good. Automated digital OSF staining assessment may provide useful endpoints for future clinical trials.
Keywords: cornea: tears/tear film/dry eye • imaging/image analysis: clinical • cornea: clinical science