Sessions

Corneal ulceration is one of the most common canine eye diseases, and the choice of therapy depends mainly on its depth. Anterior segment optical coherence tomography (OCT) shows the cornea in cross-section at micrometer resolution. Whether OCT images saved during routine examinations can be evaluated retrospectively is not well known. We aimed to establish what proportion of a routine OCT archive can be evaluated, which ulcer morphologies it shows, how measurements on archived images agree with those of the clinician, and how well two raters agree in classification. The anonymised archive of a referral veterinary ophthalmology practice in Szentes, Hungary, contained 137 OCT images (iVue, Optovue) from 54 examination dates (2023-2026); the first OCT examination of 50 eyes of 43 dogs was analysed. Each eye received a predefined evaluability status: S0, not assessable; S1, ulcer not detectable in the archived section; S2, morphology assessable but thickness not measurable; S3, morphology and thickness assessable. The morphology of the ulcers was classified into four predefined categories. Thickness was measured in ImageJ, documented as image overlays and compared with the corresponding calliper measurements recorded by the clinician. A second veterinarian of the practice independently classified 23 masked images; agreement was expressed as Cohen's kappa.

Of the 50 eyes, 3 were S0, 30 S1, 13 S2 and 4 S3. Morphology was thus assessable in 17 eyes (34 %): epithelial loss in 5, stromal loss with a smooth floor in 5, with a hyperreflective floor in 5, and descemetocele or perforation in 2. On average 2.5 OCT images were saved per examination date, compared with a median of 28 fundus photographs; sedation was documented on 2 of 40 dates. ImageJ measurements of residual thickness differed from the callipers of the clinician by 3-16 µm (3 comparisons). The raters agreed on 11 of 18 images (61 %; kappa 0.53, 95 % CI 0.26-0.78); disagreement concerned mainly superficial epithelial defects, whereas stromal defects were classified concordantly in 6 of 7 images. Dogs are mostly examined while conscious, and the clinician evaluates the live scan; the archive therefore documents only part of what was seen. In one third of the eyes the morphology was analysable retrospectively and thickness in only a few. Rater agreement was moderate, limited mainly by superficial lesions. Even so, standardised saving of sections through the deepest point of the ulcer, with scale bar and quality index, would enable prospective studies.