Purpose: To investigate the ability of three diagnostic tests: frequency-doubling technology (FDT), scanning laser polarimetry (GDx), and nerve fiber layer (NFL) photographs to distinguish normal from glaucomatous eyes.
Methods: Data were obtained in a cross-sectional, hospital clinic-based study, including one eye from each of 253 persons older than 40 years (68 normal, 94 glaucoma suspects and 91 glaucoma patients). We performed a comprehensive ocular examination, as well as static automated perimetry (Humphrey 24-2), screening FDT, GDx, optic nerve stereoscopic photographs and high-contrast NFL photographs.
Results: The following were significantly different for glaucomatous patients compared with suspects and normals: mean values of mean deviation (MD, Humphrey 24-2) and corrected pattern standard deviation (CPSD), 11 GDx indices, mean FDT testing time and missed points, and NFL graded defects (ANOVA, Mantel-Haenszel test; p = 0.0001). Using Humphrey 24-2 test results and clinical assessment as the defining features of glaucoma, we found that the optimal mix of sensitivity and specificity values were 84% and 100% for FDT (presence of any defect); 62% and 96% for GDx (The Number, cut-off value of 27); and, 95% and 82% for NFL photographs (presence of any abnormality). FDT testing took the least time to be administered.
Conclusions: The FDT had the best diagnostic performance. Neural network analysis of GDx data outperformed other elements of its software.