Glaucomatous Progression in Series of Stereo-Paired Photographs and Heidelberg Retinal Tomography Images

Abstract

Purpose: : Examining change over time as detected by automated analysis of longitudinal series of Heidelberg Retinal Tomography (HRT) images of the Optic Nerve Head (ONH) and making comparisons with the assessments, by glaucoma specialists, of change over time in stereo-paired photography of the same ONHs.

Methods: : Baseline and follow-up stereo-paired photographs and their corresponding HRT series of 107 eyes from 62 patients were selected from those of 336 eyes of 168 patients^1. The principal criteria of sufficiently long, good quality HRT series (at least 6 follow-up visits over at least 6 years duration) and the follow-up photograph being taken contemporary to or after the last HRT image acquisition was used for selection. Statistic Image Mapping (SIM, Patterson et al 2005 IOVS 46: 1659-67.), Topographic Change Analysis (TCA) and linear regression of optic disc Rim Area (RA) were applied to all 107 HRT series. Type and location of progressive damage were determined in the stereo-paired photographs as was sectoral location of change in TCA and SIM defined by the location of centre of largest flagged cluster.

Results: : Using the expert assessments of progression in the stereo-paired photographs as a reference standard, HRT analysis had sample specificity and sensitivity of 83% and 34% (SIM), 73% and 36% (TCA), 82% and 40% (linear regression of RA). Sectoral location agreement between SIM and TCA was 67% and for SIM, TCA and RA change at 53%. Out of those eyes flagged as progressing the type of change, as assessed from stereo-photographs most frequently detected by SIM, TCA and linear regression of RA was notching (29%), rim thinning (38%) and rim thinning (42%) respectively.

Conclusions: : The ability of automated techniques in detecting similar significant glaucomatous progressive damage as measured by glaucoma specialists in stereo-paired photographs is poor but the agreement between methods of automated HRT progression analysis is good. This study suggests that analyses of HRT series and expert assessed stereo-paired photographs provides largely independent information about glaucomatous change. The results illustrate the difficulty of comparing methods to detect structural progression in glaucoma in the absence of a gold-standard.1. Mansberger et al (2007). Invest Ophthalmol Vis Sci 48: E-Abstract 3338.

Keywords: optic disc • imaging/image analysis: clinical • microscopy: confocal/tunneling