A clinical study of perimetric probability maps.

Perimetric probability maps depict visual field results in terms of the frequency with which the measured findings are seen in a normal population. We tested clinically the importance of the model of the normal visual field used to calculate such maps. Forty-one eyes of 41 normal subjects and 58 eyes of 46 glaucomatous patients were studied. Probability maps were calculated by means of two different models of the normal visual field. The first model assumed gaussian threshold distributions with constant variability across the field. The second used empirically determined nongaussian location-dependent threshold distributions. Probability maps using the empiric model allowed better separation between glaucomatous and normal eyes, and the number of significant points in normal subjects was in better agreement with the theoretically expected number. The gaussian model yielded an unacceptably high frequency of significant points in normal fields, particularly in the midperiphery. The clinical usefulness of perimetric probability maps depends critically on the choice of normal visual field model.