Comparison of diagnostic performance and fixation control of two automated perimeters

Abstract

Background: The Humphrey perimeter and its Statpac (analysis programs have been widely used and studied. Another statistical analysis program, FieldView, is used with the Dicon perimeter. The purpose of this study was to compare the diagnostic performance of the two perimeters and their statistical analysis packages.

Methods: Twenty-three normal subjects (age range, 27 to 83 years) and 31 patients with glaucoma or cerebrovascular disease (age range, 28 to 87 years) experienced in automated perimetry were examined using the Dicon and the Humphrey perimeters.

Results: The total number of significant points identified on the Humphrey total deviation probability maps was in close agreement with statistical expectations, while the Dicon total deviation probability maps yielded significantly more false-positive defects than expected for normals. Fixation loss ratios were almost twice as high with the Dicon perimeter (mean, 16%) as compared with the Humphrey perimeter (mean, 9%). The Humphrey perimeter was more reliable than the Dicon in measuring the defect depth of the physiological blind spot.

Conclusion: The Dicon perimeter appears to yield excessive false-positive findings in normal subjects, resulting in poor sensitivity/specificity combinations, while at the same time failing to properly measure defect depth in scotomas.