Automatic Segmentation and Visualization of Choroid in OCT with Knowledge Infused Deep Learning
The choroid provides oxygen and nourishment to the outer retina thus is related to the pathology of various ocular diseases. Optical coherence tomography (OCT) is advantageous in visualizing and quantifying the choroid in vivo. (1) The lower boundary of the choroid (choroid-sclera interface) in OCT is fuzzy, which makes the automatic segmentation difficult and inaccurate. (2) The visualization of the choroid is hindered by the vessel shadows from the superficial layers of the inner retina. In this paper, we propose to incorporate medical and imaging prior knowledge with deep learning to address these two problems. We propose a biomarker infused global-to-local network for the choroid segmentation. It leverages the choroidal thickness, a primary biomarker in clinic, as a constraint to improve the segmentation accuracy. We also design a global-to-local strategy in the choroid segmentation: a global module is used to segment all the retinal and choroidal layers simultaneously for suppressing overfitting and providing global structure information, then a local module is used to refine the segmentation with the biomarker infusion. To eliminate the retinal vessel shadows, we propose a pipeline that firstly use anatomical and OCT imaging knowledge to locate the shadows using their projection on the retinal pigment epithelium layer, then the contents of the choroidal vasculature at the shadow locations are predicted with an edge-to-texture generative adversarial inpainting network. The experiments show our method outperforms the existing methods on both the segmentation and shadow elimination tasks. We further apply the proposed method in a clinical prospective study for understanding the pathology of glaucoma by detecting the structure and vascular changes of the choroid related to the elevation of intra-ocular pressure.