To compare the ability of Heidelberg Retina Tomograph (HRT) 2 and the new version, HRT3, to discriminate between glaucoma and non-glaucoma eyes and to evaluate the most useful parameters for glaucoma detection.

Fifty-two healthy eyes, 62 glaucoma-suspect eyes, and 172 glaucoma eyes (a total of 286 eyes) were studied retrospectively. The discrimination capabilities of healthy and glaucomatous eyes were compared using areas under the receiver operating characteristics (AROCs) curves. Agreement between classifications was defined by κ analysis.

The stereometric parameters that best discriminated between glaucoma and non-glaucoma eyes were cup shape measure at HRT2 (areas under the ROCs curves (AUC) 0.784) and global probability score (GPS) at HRT3 (AUC 0.718). The stereometric parameters that best discriminated between normal and non-normal eyes were cup shape measure at HRT2 (AUC 0.785) and GPS at HRT3 (AUC 0.696). The sensitivity was 95% at HRT3 and 91% at HRT2, and specificity was 61% at HRT2 and 59% at HRT3.

The glaucoma-discriminating ability of the HRT3 software is similar to that of HRT2; however, HRT3 provided results without the need for subjective operator input.