Abstract
Purpose: :
To demonstrate a new method of retinal pigment epithelium (RPE) segmentation by polarization sensitive optical coherence tomography (PS-OCT).
Methods: :
A spectral domain PS-OCT system was developed that allows to simultaneously record 2- and 3-dimensional data sets of backscattered intensity, retardation, and optic axis orientation. More than 100 eyes of subjects with macular diseases and 10 eyes of healthy individuals were scanned. In previous work we found that the RPE, contrary to other retinal layers, depolarizes backscattered light (i.e. it scrambles its polarization state). This polarization scrambling causes a broad distribution of retardation values obtained from the RPE and therefore can be used as a tissue specific, intrinsic contrasting mechanism. The retardation signal in each PS-OCT B-scan was analyzed within a moving window (size: 20 x 8 pixels). For each window position, a histogram of retardation values was calculated. If the width of the retardation distribution was larger than a user defined threshold, the corresponding area was identified and color-coded as RPE tissue.
Results: :
The algorithm worked well in most cases, being able to segment the RPE in normal eyes as well as in eyes with drusen, RPE atrophy, pseudovitelliform macular dystrophy, teleangiectasia, macular holes, and others. In case of RPE atrophy, the lateral extension of the lesions can readily be displayed and measured. Only in cases with poor signal quality (e.g. media opacities) the polarization data are unreliable and the algorithm cannot be used.
Conclusions: :
PS-OCT provides a new way for identification and segmentation of the RPE. Contrary to conventional, intensity based segmentation algorithms our algorithm exploits an intrinsic tissue property, depolarization. It is therefore less vulnerable to discontinuities of tissue contours within the image that might be caused by real tissue defects like atrophies and holes, or by image artifacts like vessel shadows or motion artifacts. PS-OCT might improve diagnosis and precise follow-up of RPE-related diseases.
Keywords: imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • retinal pigment epithelium • image processing