Abstract
Abstract: :
Purpose: To reduce variation between sequential mean topography images of the optic nerve head (ONH). Methods: During a Heidelberg Retinal Tomograph (HRT) scan, three stacks of confocal images are obtained, from which three–dimensional (3D) images of the ONH are generated. Current HRT software generates three topography images which are combined to form a mean topography image. We propose a novel method whereby each stack of confocal images is aligned to compensate for translational and rotational disparities which occur due to fixation errors during a scan. The three image stacks are then globally aligned in three translations and three rotations, using a sub–pixel alignment algorithm, and the mean image stack is generated. The mean topography image is then generated from the mean image stack. The current method used by the HRT software to generate the mean topography images was simulated. Inter–(mean) topography variability was calculated as the mean standard deviation of height measures in two sets of three mean topographies recorded by a normal subject. The differences in inter–topography variability between the simulated HRT method and the new method were calculated. Results: The new method yielded a 13% reduction in mean standard deviation of height measures in the example subject as compared to the simulated method used by the HRT software. The resulting mean standard deviation of height measures were 14.9 µm and 17.1 µm, respectively, for the new and current methods. The generation of a mean topography image using the new method takes approximately seven minutes on a Pentium IV 2GHz processor. Conclusions: The new method reduced the variation between mean topography images, compared to a simulation of the current software. The benefit of this technique will be evaluated by the effect of improved reproducibility of topography images on the repeatability of HRT stereometric parameters.
Keywords: imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • motion–3D • image processing