Segmentation of Corneal Optical Coherence Tomography Images Using Graph Search and Radon Transform

摘要

Various common corneal eye diseases, such as dry eye, Fuchs endothelial dystrophy, Keratoconus and cornealgraft rejection, can be diagnosed based on the changes in the thickness of corneal microlayers. Optical CoherenceTomography (OCT) technology made it possible to obtain high resolution corneal images that show the micro-layered structures of the cornea. Manual segmentation is subjective and not feasible due to the large volumeof obtained images. Existing automatic methods, used for segmenting corneal layer interfaces, are not robustand they segment few corneal microlayer interfaces. Moreover, there is no large annotated database of cornealOCT images, which is an obstacle towards the application of powerful machine learning methods such as deeplearning for the segmentation of corneal interfaces. In this paper, we propose a novel segmentation method forcorneal OCT images using Graph Search and Radon Transform. To the best of our knowledge, we are the firstto develop an automatic segmentation method for the six corneal microlayer interfaces. The proposed methodinvolves a novel image denoising method and an inner interfaces localization method. The proposed methodwas tested on 15 corneal OCT images. The images were randomly selected and manually segmented by twooperators. Experimental results show that our method has a mean segmentation error of 3:87 - 5:21 pixels (i.e.5:81 - 7:82μm) across all interfaces compared to the segmentation of the manual operators. The two manualoperators have mean segmentation difference of 4:07 - 4:71 pixels (i.e. 6:11 - 7:07μm). The mean running timeto segment all the corneal microlayer interfaces is 6:66 - 0:22 seconds.