Purpose: To develop a clinically specific strategy for optic nerve head (ONH) surface change detection within longitudinally acquired confocal scanning laser tomographic (CSLT) images.
Design: Experimental study.
Participants and/or controls: Twelve monkeys, each with one glaucomatous and one contralateral normal eye.
Main outcome measures: Detection of ONH surface change within the CSLT images of each monkey's glaucomatous eye.
Methods: Six 10 and six 15 CSLT images and four stereo optic disc photographs were obtained from both eyes of 12 monkeys on 3 separate days (imaging sessions) and then every 2 weeks after laser to one eye (study eye) to elevate intraocular pressure. For each set of 10 and 15 images per eye (and for each of the 7-25 imaging sessions per monkey), 37 CSLT parameters were calculated. Univariate change required an individual parameter to change in excess of its analysis of variance-determined minimum detectable change (MDC) value. Multivariate change required groups of three parameters, considered together, to demonstrate significant change as determined by a multivariate analysis of variance. The rate of false-positive change detection for each individual parameter and for a group of three-parameter combinations was determined using both a one-in-a-row strategy (change at a single session) and a two-in-a-row strategy (change at two successive sessions) within the postlaser images of the 12 unchanging contralateral normal eyes. Change detection within the study eye images was then assessed for only those individual parameters and three-parameter combinations that were clinically specific (i.e., showed less than 10% false-positive change detection in the normal eyes).
Results: A total of 36 prelaser (three per monkey) imaging sessions and 158 postlaser (4-22/monkey) imaging sessions was performed. Clinically specific change detection (low rate of false positives) was achieved only with the two-in-a-row strategy. Overall, multivariate ONH surface change detection performed best; the best-performing three-parameter combination detected only 8 change events (4 onset and 4 progression) in 139 imaging sessions within the postlaser imaging sessions of the contralateral normal eyes and a total of 47 change events (11 onset and 36 progression) within the postlaser imaging sessions of the 12 study eyes. Counterintuitive (anterior) change occurred in most parameters within the late postlaser imaging sessions of the study eyes followed to end-stage damage.
Conclusions: Clinically specific detection of the onset and progression of glaucomatous ONH surface change is possible within longitudinally acquired CSLT images.