Abstract
Purpose: :
To evaluate the diagnostic accuracy of the Heidelberg Retina Tomograph 3 (HRT 3) for glaucoma detection in an elderly population.
Methods: :
1952 participants of the 10-year follow-up Blue Mountains Eye Study were studied. Optic nerve head scans were performed using HRT 2 hardware. All data were converted to HRT 3 software for analysis. Open angle glaucoma was diagnosed independently from optic disc photographs and Humphrey 24-2 visual fields. Sensitivity and specificity were used to evaluate the diagnostic ability of HRT 3. The Moorfields Regression Analysis (MRA) and Glaucoma Probability Score (GPS) were subdivided into result 1 (MRA1 and GPS1) with "borderline" outcomes as negative and results 2 (MRA2 and GPS2) with "borderline" outcomes as positive. The diagnostic accuracy of combined classification of MRA and GPS was further evaluated.
Results: :
HRT scans were acquired in 1644 of 1952 participants. Data from 1640 participants was successfully converted to HRT 3. Weighted Kappa was 0.317 for agreement between classifications of MRA and GPS. For analyses by person, sensitivity and specificity for MRA1 were 81.7% and 78.0% and MRA2 were 92.5% and 51.0% respectively. GPS1 had 11.6% increase in sensitivity and 14.7% decrease in specificity compared to MRA1. Compared with MRA2, GPS2 sensitivity only improved by 3.1% but specificity dropped by 18.8%. Similar differences were found for analyses by eye. The combined classification of MRA and GPS gave the sensitivity of 86.0% and specificity of 73.2% at person’s level and sensitivity of 81.5% and specificity of 82.0% at eye’s level. Older age and larger optic disc size were associated with lower MRA and GPS specificity. But the size of this effect was reduced compared with the HRT 2. Among all optic disc parameters, larger cup size was found to be the dominant predictor for false positive outcome for MRA and GPS.
Conclusions: :
Compared with HRT 2, HRT 3 diagnostic tests showed improved sensitivity, but at the expense of specificity in this population-based setting. The combined MRA and GPS classification provided better diagnostic accuracy than either MRA or GPS alone.
Keywords: imaging/image analysis: clinical • optic disc • clinical (human) or epidemiologic studies: prevalence/incidence