Corneal Segmentation Analysis Increases Glaucoma Diagnostic Ability of Optic Nerve Head Examination, Heidelberg Retina Tomograph’s Moorfield’s Regression Analysis, and Glaucoma Probability Score

Abstract

Purpose. To study whether a corneal thickness segmentation model, consisting in a central circular zone of 1 mm radius centered at the corneal apex (zone I) and five concentric rings of 1 mm width (moving outwards: zones II to VI), could boost the diagnostic accuracy of Heidelberg Retina Tomograph’s (HRT’s) MRA and GPS. Material and Methods. Cross-sectional study. 121 healthy volunteers and 125 patients with primary open-angle glaucoma. Six binary multivariate logistic regression models were constructed (MOD-A1, MOD-A2, MOD-B1, MOD-B2, MOD-C1, and MOD-C2). The dependent variable was the presence of glaucoma. In MOD-A1, the predictor was the result (presence of glaucoma) of the analysis of the stereophotography of the optic nerve head (ONH). In MOD-B1 and MOD-C1, the predictor was the result of the MRA and GPS, respectively. In MOD-B2 and MOD-C2, the predictors were the same along with corneal variables: central, overall, and zones I to VI thicknesses. This scheme was reproduced for model MOD-A2 (stereophotography along with corneal variables). Models were compared using the area under the receiver operator characteristic curve (AUC). Results. MOD-A1-AUC: 0.771; MOD-A2-AUC: 0.88; MOD-B1-AUC: 0.736; MOD-B2-AUC: 0.845; MOD-C1-AUC: 0.712; MOD-C2-AUC: 0.838. Conclusion. Corneal thickness variables enhance ONH assessment and HRT’s MRA and GPS diagnostic capacity.