Abstract
Purpose: :
To describe the characteristics of eyes that were classified correctly and incorrectly most often by general ophthalmologists in the European Optic Disc Assessment Trial.
Methods: :
Simultaneous stereoscopic optic disc photographs of one eye each of 40 healthy subjects and 48 glaucoma patients were obtained. Measurements with scanning laser polarimetry with variable corneal compensation (SLP-VCC; GDx VCC) and confocal scanning laser ophthalmoscopy (CSLO; Heidelberg Retina Tomograph [HRT]) were also performed. From 11 European countries, 244 general ophthalmologists classified the photographs as normal or glaucomatous. We determined the characteristics of the 6 eyes that were misclassified most often by the ophthalmologists as well as the 6 eyes that were correctly classified most often.
Results: :
Ophthalmologists had a mean (SD) sensitivity, specificity, and overall accuracy of 74.7% (10.6), 87.5% (10.9), and 80.5% (6.8), respectively. The overall accuracy of the GDx VCC by means of the NFI was 93.2%. For the HRT, the overall accuracy was 89.8% by means of the Bathija linear discriminant function (LDF) and 87.5% by means of the Moorfields regression analysis (MRA). For both the correctly and incorrectly classified eyes, 5 of the 6 eyes were glaucomatous. The mean (SD) Mean Deviation (MD) and Pattern Standard Deviation (PSD) were -1.98 dB (0.92) and 3.86 dB (1.58) in the worst classified eyes, respectively, and -12.79 dB (8.00) and 11.43 dB (5.03), respectively, in the best classified eyes. Differences in MD and PSD between these 2 groups were statistically significant (p<0.01). The mean (SD) size of the optic disc as measured with CSLO was 1.69 mm2 (0.18) and 2.07 mm2 (0.54) in the worst and in the best classified eyes, respectively (p=0.13). In the 6 eyes that were misclassified most often, 5 were classified correctly by the GDx VCC NFI, 4 were classified correctly by the HRT Bathija LDF, and 1 was classified correctly by the HRT MRA. In the 6 eyes that were classified correctly most often, all were classified correctly by the GDx VCC NFI, all but one were classified correctly by the HRT Bathija LDF, and all were classified correctly by the HRT MRA.
Conclusions: :
In glaucoma, optic discs of smaller size and with milder glaucomatous loss are most difficult to be classified correctly by general ophthalmologists. On average, automated analysis of measurements with imaging techniques classify these eyes better. We think that imaging techniques such as the GDx VCC and HRT may provide additional information to the ophthalmologist in diagnosing glaucoma, especially in mild cases.
Keywords: optic disc • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound)