Abstract
Abstract: :
Purpose: To determine the effect of inter- and intra-observer variability in contour line placement on confocal scanning laser tomography and scanning laser polarimeter parameters. Methods: From a database of patients attending a longitudinal glaucoma study a single image with the Heidelberg Retina Tomograph (HRT) and Laser Diagnostic scanning laser polarimeter (GDx) was selected for 20 normals, 20 ocular hypertensives and 20 glaucoma patients. Five trained operators defined the disc margins and 3 repeated the measurements after ≷1 week. Six parameters were selected for each instrument and intra-class correlation coefficients (ICC) computed. Diagnostic category was also compared using the HRT discriminant analysis function and recommended interpretation of the GDx 'number'. Results: The table shows the ICC for each parameter. Inconsistency in diagnostic category between the 5 observers was found in 20 (HRT) and 15 (GDx) cases. There was no tendency for a specific group of patients be misclassified with either instrument (Chi-squared test). Intra-observer parameter variability for the 3 observers was: HRT (mean ICC (range)) 0.964 (0.920 - 0.988), 0.945 (0.853 - 0.982) and 0.687 (0.557 - 0.772) and GDx 0.986 (0.953 - 0.996), 0.990 (0.976 - 0.999) and 0.968 (0.907 - 0.999). Kappa statistic for agreement of the diagnostic category was: 0.897, 0.798 and 0.564 (HRT) and 0.827, 1.00 and 0.779 (GDx). Conclusion: Although good agreement was found between operators with both instruments, the diagnostic categorisation varied with contour line placement. Also, the range of intra-observer variability shows the importance of correctly establishing the disc margin if the instruments are used as clinical tools.
Inter-observer agreement of HRT and GDx parameters
Keywords: 430 imaging/image analysis: clinical • 498 optic disc