Automated gonioscopy photography for iridocorneal angle grading

Abstract

PurposeThis study aims to assess inter‐device and inter‐observer agreement of iridocorneal angle grading using automatic gonio‐photography and dynamic gonioscopy.MethodsCross sectional single‐center observational study. Consecutive patients from Glaucoma Clinic were recruited to perform two types of gonioscopy. Manual gonioscopy (MG) was performed using a dynamic gonio lens, while the automatic Gonio‐Photography (AG) was done using a contact‐gonioscope prototype (from NIDEK). AG images were graded by two masked observers (glaucoma specialist and resident). Criteria for classification was similar in both methods. An eye was classified with angle closure when the posterior trabecular meshwork couldn't be seen in two or more quadrants. Inter‐device agreement (AG vs MG) was ascertained by κ statistic and comparison of area under receiver operating characteristic curves (AUC). Agreement among raters was ascertained by Fleiss’ κ statistic and intraclass correlation coefficient (ICC).ResultsForty‐eight eyes were analyzed. Mean age was 63 ± 16 years. Angle closure was detected in 11 eyes (23.4%) with MG in comparison with 2 eyes (4.3%) using AG images grading. The agreement for angle closure diagnosis between MG and AG was poor (κ=0.09 ± 0.10; p = 0.18). The AUC for detecting eyes with gonioscopic angle closure showed poor accuracy between automated and manual methods (AUC=0.53 ± 0.05, confidence interval (CI) 95%: 0.44‐0.62). The ICC for iridocorneal angle grading between the two raters was 0.34 [CI 95%: 0.10‐0.51] and Fleiss’ kappa coefficient for opening status was 0.13 [CI 95%: 0.04‐0.30].ConclusionsOnly slight agreement was found when comparing automated with manual dynamic gonioscopy and also when comparing two raters. Manual dynamic gonioscopy is the gold standard and further technological development is desired for promising automated gonioscopy.