A new statistical approach for quantifying change in series of retinal and optic nerve head topography images

To describe and evaluate new statistical techniques for detecting topographic changes in series of retinal and optic nerve head images acquired by scanning laser tomography (Heidelberg Retinal Tomograph [HRT]; Heidelberg Engineering, Heidelberg, Germany).

Proven quantitative techniques, collectively referred to as statistic image mapping (SIM), are widely used in neuroimaging. These techniques are applied to HRT images. A pixel-by-pixel analysis of topographic height over time yields a statistic image that is generated by using permutation testing, derives significance limits for change wholly from the patient's own data, and removes the need for reference data sets. These novel techniques were compared to the Topographic Change Analysis (TCA super-pixel analysis) available in the current HRT software, by means of an extensive series of computer experiments. The SIM and TCA techniques were further tested and compared to linear regression of rim area (RA) against time, in real longitudinal HRT series of eyes of 20 normal subjects and 30 ocular hypertensive (OHT) patients that were known to have converted to glaucoma, on the basis of visual field criteria.

Computer simulation indicated that SIM has better diagnostic precision at detecting change. In the real longitudinal series, SIM flagged false-positive structural progression in two (10%) of normal subjects, whereas TCA identified three (15%), and linear regression of RA against time identified two (10%). SIM identified 22 (73%) of the OHT converters as having structural progression, whereas the TCA and linear regression of RA against time each identified 16 (53%) over the course of the follow-up.

SIM has better diagnostic precision in detecting change in series of HRT images when compared to current quantitative techniques. The clinical utility of these techniques will be established on further longitudinal data sets.