Diagnostic value of multifocal VEP using cross-validation and noise reduction in glaucoma research.
BACKGROUND: The diagnostic value of multifocal visual evoked potentials (mf VEP) in glaucoma research is still under debate. Several previous studies proclaim it to be a useful tool for clinical applications, but according to other studies, different problems (low specificity, poor records, and interindividual variation) still retard its clinical use. The aim of the present study was to examine whether the mf VEP data obtained with the RETIscan system are appropriate for formulating a classification rule for glaucoma. METHOD: We examined and evaluated 65 eyes in 38 advanced glaucoma patients and 27 normal subjects, using four occipital gold cup electrodes (cross layout) for bipolar recording and a CRT monitor (display diameter 60 degrees, chequerboard pattern reversal, 60 segments in dartboard layout) for stimulation. In each case, eight cumulative measurements (77 s each) were made. The data of the 60 segments were cross-correlated with a RETIscan-internal VEP norm ("VEP finder"), combined in 16 sectors, and evaluated via the classification technique "double-bagging" and the Wilcoxon U-test. RESULTS: In three out of the 16 sectors, the VEP amplitudes of the patients were significantly reduced (Wilcoxon U-test). Applying double-bagging on the cross-correlated data (with VEP finder) resulted in a sensitivity of 75% and a specificity of 71%, and the estimated misclassification rate was 27%. For uncorrelated data (without VEP finder), the same analysis achieved a sensitivity of about 60% and a specificity of 40%. CONCLUSIONS: Estimated sensitivity and specificity suggest that by using the RETIscan system for recording, a classification of the VEP data--i.e. a separation between normal and glaucoma subjects--is possible.