Purchase this article with an account.
F. K. Horn, W. Adler, R. Laemmer, G. Michelson, A. G. Juenemann, B. Lausen; Glaucoma Detection With FDT-Screening and HRT-Measurements Using Classification Algorithms in Training and Test Populations. Invest. Ophthalmol. Vis. Sci. 2007;48(13):1633.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To develop classification rules for combined analysis of data from frequency doubling technology (FDT) perimetry and results from HRT measurements and to test the diagnostic value of this newly developed algorithm in independent patient groups.
The training sample for determination of new classifiers included 174 participants of the Erlangen glaucoma registry: 54 patients with "preperimetric" open angle glaucoma (glaucomatous optic disc atrophy and no visual field defect in standard white on white perimetry), 73 patients with "perimetric" open angle glaucoma (glaucomatous optic disc atrophy and visual field defect in standard perimetry), and 47 control subjects. To study the classification with respect to their missclassification error rates, an independent second set of 267 patients was recruited from the Erlangen university eye hospital (36 preprimetric, 94 perimetric OAG and 137 control subjects). All subjects had FDT screening (protocol: C-20-5) and Heidelberg Retina Tomograph (HRT) scanning of the optic disc as well as conventional perimetry and evaluation of fundus photographs. Criteria for exclusion: media opacities, diabetes, subjects younger than 35 years or older than 65 years. One eye of each patient and control subject entered the statistical evaluation. Analysis for combination of FDT scores and HRT results used linear discriminant analysis and double-bagging machine-learning algorithms including information of global HRT standard variables and FDT quadrants.
A combination of FDT perimetry and HRT yielded better diagnostic results than each of the diagnostic instruments separately. In the training data set the area under the ROC-curve of the combined classifier was 87.9% compared to 76.7% for HRT and 83.8% for FDT perimetry. The performance of classifiers using combination of both instruments is also superior to the individual instruments in the independent test set. (area under ROC of combined classifier: 84.9%, HRT: 75.0%, FDT: 80,6%).
Joint analysis of data from HRT measurements and FDT perimetry indicate that a combination of both techniques might be helpful in a glaucoma screening setup. This could be shown in groups of patients from an ongoing case-control study as well as in independent populations of an outpatient department.
This PDF is available to Subscribers Only