Abstract
Purpose: :
The CirrusTM HD-OCT 4.0 software has a Macular Change Analysis(MCA) module that incorporates image registration to allow the user to evaluate changes between visits. The purpose of this paper is to evaluate the benefit of image registration in MCA to compare thickness measurements.
Methods: :
We retrospectively analyzed macula scans from 17 normal subjects, collected as part of an in-house reproducibility study. On each day, 3 separate Macula 200x200 scans were obtained giving a total of 9 scans per subject. Analysis of variance (ANOVA) was used to determine the repeatability of the central subfield thickness with no registration, with a translation only registration, and a rigid registration (translation plus rotation) between the scans. The analysis was also done on mean thickness from super-pixels (4x4 pixels) over the entire image. In order to compare different transformation models, we looked at the max, min, median and average values for the repeatability for the super pixels.
Results: :
The rigid registration used in Cirrus 4.0 gives the best results followed by translation only and then no registration. Table 1 shows the variance components from the ANOVA, and the repeatability standard deviation (square root of random variance). Table 2 summarizes the results of the super pixel analysis. The rigid transformation again produces better overall repeatability.
Conclusions: :
Image registration in Macular Change Analysis module of CirrusTM HD-OCT 4.0 software helps to improve the repeatability of Cirrus HD-OCT thickness measurements. The rigid transformation model, which is used in the software, gives the best repeatability among the transformation models considered.
Keywords: imaging/image analysis: clinical