Objective: To evaluate and compare the diagnostic accuracy of the Heidelberg retina tomograph (HRT) version 3 with that of glaucoma specialists using stereophotography in discriminating between normal eyes and patients with early glaucomatous visual field loss.
Methods: A total of 105 eyes of 105 individuals were prospectively and consecutively recruited. The sample comprised 51 normal and 54 early glaucomatous eyes, as defined by intraocular pressure and standard automated perimetry results, regardless of optic disc appearance. Receiver operating characteristic (ROC) curves were plotted for the HRT3 parameters and a linear discriminant function (LDF) developed in our hospital. Best sensitivity-specificity pairs were compared between the HRT3 parameters, with the highest areas under the ROC curve (AUCs) and evaluation of optic disc stereophotographs. Agreement between methods for measuring vertical cup-to-disc ratio was evaluated with the Bland-Altman plot.
Results: The average visual field mean deviation was -2.90 dB. The HRT3 parameters with the largest AUCs were our LDF (0.900), rim volume (0.883), and vertical cup/disk ratio (0.880), with no significant differences between these parameters. Sensitivity-specificity pairs were 79.6-100% (clinical evaluation), 83.3-86.3% (our LDF), 64.8-96.1% (final glaucoma probability score), and 68.5-90.2% (global Moorfields regression analysis).
Conclusions: The diagnostic accuracy for differentiating normal eyes from those with early visual field defects was similar between clinical evaluation of the optic disc and evaluation with the HRT3. The use of our LDF increased the sensitivity-specificity balance with respect to the HRT-provided parameters. The diagnostic accuracy of the HRT classifications was comparable to that of an experienced glaucoma specialist.