Segmentation of Retinal Lesions by Polarization Sensitive Optical Coherence Tomography

Abstract

Purpose: : To demonstrate new algorithms based on polarization sensitive optical coherence tomography (PS-OCT) for segmenting retinal lesions like retinal pigment epithelium (RPE) atrophies, drusen, hard exudates and subretinal fluid.

Methods: : An experimental spectral domain PS-OCT system was used that allows to simultaneously record 2- and 3-D data sets of backscattered intensity and various polarization parameters in the human retina. We previously have shown that the RPE depolarizes backscattered light; this effect can be used to segment the RPE based on the parameter "degree of polarization uniformity" (DOPU). We demonstrate the use of this method to segment and quantify RPE atrophies. Furthermore, we expand on this method and developed new algorithms that use the segmented RPE as a backbone for locating other boundaries like Bruch's membrane or the photoreceptors by intensity based segmentation methods. These layers are then used to quantify area and volume of drusen and subretinal fluid. Hard exudates are also segmented by their depolarizing properties. Data sets of ~ 60 patients with geographic atrophies, drusen, hard exudates, or intraretinal fluid are analyzed.

Results: : Segmentation of RPE atrophies worked well and showed a good correlation with autofluorescence images. Drusen segmentation worked well in most cases, with some exceptions in large, confluent drusen. Segmentation of hard exudates in patients with diabetic retinopathy showed a good correlation with fundus photography. First results in two patients with subretinal fluid demonstrate the potential of the method for quantifying fluid volume.

Conclusions: : Segmentation of retinal lesions like RPE atrophies, drusen, hard exudates, and intraretinal fluid by PS-OCT based algorithms are an interesting alternative to conventional algorithms based on OCT intensity data. The segmentation of the RPE by the tissue specific DOPU parameter provides a reliable backbone for segmentation of adjacent structures. PS-OCT might improve diagnosis and precise follow-up of RPE-related diseases.

Keywords: imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • image processing • imaging/image analysis: clinical