Purchase this article with an account.
A. Ferreras, A. B. Pajarin, L. E. Pablo, P. Calvo, B. Monsalve, P. Fogagnolo, M. Figus; Logistic Regression Analysis for Glaucoma Diagnosis Using Spectral Domain Optical Coherence Tomography. Invest. Ophthalmol. Vis. Sci. 2010;51(13):225.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To design and evaluate the diagnostic ability of a linear discriminant function (LDF) based on the peripapillary retinal nerve fiber layer (RNFL) thickness obtained using spectral domain optical coherence tomography (OCT) for discriminating between healthy eyes and eyes with glaucomatous visual field loss.
Eighty-eight healthy subjects and 65 patients with open-angle glaucoma were consecutive and prospectively selected. All of them underwent imaging with the Optic Disc Cube 200 x 200 scan protocol of Cirrus HD OCT (Carl Zeiss Meditec, Dublin, Ca). Only one eye was randomly chosen. Left eye data were converted to a right eye format. The LDF was calculated according to the stepwise logistic regression results of the mean RNFL thickness at the 12 clock-hour positions and in the four quadrants. The receiver operating characteristic (ROC) curves were plotted for the parameters included in the software of the OCT, and compared with the LDF. Differences between the areas under the ROC curves were tested using the Hanley-McNeil method.
The obtained learning classifier was: LDF = 16.267 - (0.082 x inferior quadrant) - (0.068 x superior quadrant). The areas under the ROC curve were 0.954 for the LDF, 0.938 for the inferior quadrant thickness, and 0.931 for the average thickness. There were no significant differences between these values.
Compared with the OCT-provided parameters, the LDF had the highest sensitivity at 95% fixed specificity to discriminate between normal and glaucomatous eyes. Learning classifiers may help to increase diagnostic ability of OCT.
This PDF is available to Subscribers Only