A comparison of the glaucoma probability score to earlier heidelberg retina tomograph data analysis tools in classifying normal and glaucoma patients

Abstract

Purpose: To compare the performance of the Glaucoma Probability Score (GPS), Mikelberg linear discriminant function (LDF), Burk LDF, and Moorfield regression analysis (MRA) in classifying optic disc images acquired from normals and glaucoma patients with the Heidelberg Retina Tomograph (HRT).

Patients and methods: This is a retrospective comparative study of 110 eyes of 110 subjects clinically categorized as glaucoma or normal. Topographic images of the optic nerve head were obtained from HRT. Data analysis of the HRT images were carried out using GPS, Mikelberg LDF, Burk LDF, and MRA. Diagnostic performances and agreement in classification between the data analysis tools were calculated and compared for GPS, Mikelberg LDF, Burk LDF, and MRA.

Results: The highest specificity but lowest sensitivity value was obtained using the Burk LDF. The highest sensitivity but lowest specificity was obtained using the GPS. The GPS and MRA were however similar in performance. Sensitivity and specificity values of the GPS and other LDFs were affected by disc size.

Conclusions: The GPS compared similarly with the MRA without the need of additional contour line placements. Disc size is still an important factor in classification by the GPS and the other LDFs.