Purpose:
The choroidal thickness provides valuable information for early diagnosis of eye diseases. Until recently, the chorio-scleral interface (CSI) segmentation of an optical coherence tomography (OCT) image is achieved by using only the intensity (back scattering) information. Although this method would be useful, it is not fully based on histological evidence. Our research aims at developing an automated CSI segmentation algorithm based on the birefringence properties of the choroid and sclera detected by polarization sensitive OCT (PS-OCT).
Methods:
The choroid is known to have low birefringence, while the sclera is highly birefringent because of its high concentration of collagen. Hence, they could be distinguished by phase retardation tomography obtained by PS-OCT because the phase retardation is sensitive to the birefringence. In our algorithm, the CSI is defined as the cross point of depth-oriented linear fitting lines of choroidal and scleral phase retardations. Six eyes of 6 subjects without marked posterior disorder were involved in this study. A 5-mm region centered at the fovea was scanned by PS-OCT. Both intensity and phase retardation tomographies were obtained. A custom-made PS-OCT based segmentation algorithm were applied to these tomographies.
Results:
Fig. 1(a) and 1(b) show the CSI segmentation result overlaid on intensity (a) and phase retardation (b) tomographies. The birefringence property difference between choroid and sclera can be visualized as a difference in depth-oriented phase retardation gradient. In 4 of the 6 subjects, the automatically segmented CSIs were similar to those of manually obtained with the intensity tomographies. However, unsmooth CSI appeared in some regions as indicated by arrows in Fig 1(a) and 1(b). This irregular shape of CSI coincided with local high phase retardation and may suggest real weaving CSI structure that have been ignored by intensity based CSI segmentation.
Conclusions:
A phase retardation tomography based algorithm was developed for CSI detection. This approach challenges the intensity based segmentation, providing a fully automated method with histological relevance for CSI segmentation.
Keywords: choroid • image processing • imaging/image analysis: non-clinical