Automatic Segmentation and Visualization of Choroid in OCT with Knowledge Infused Deep Learning

Huihong Zhang,Jianlong Yang,Kang Zhou,Fei Li,Yan Hu,Yitian Zhao,Ce Zheng,Xiulan Zhang,Jiang Liu

DOI: https://doi.org/10.1109/jbhi.2020.3023144

IF: 7.7

2020-12-01

IEEE Journal of Biomedical and Health Informatics

Abstract: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. However, its application in the study of the choroid is still limited for two reasons. (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 (Bio-Net) for the choroid segmentation, which not only regularizes the segmentation via predicted choroid thickness, but also leverages a global-to-local segmentation strategy to provide global structure information and suppress overfitting. For eliminating the retinal vessel shadows, we propose a deep-learning pipeline, which firstly 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 results show our method outperforms the existing methods on both tasks. We further apply the proposed method in a clinical prospective study for understanding the pathology of glaucoma, which demonstrates its capacity in detecting the structure and vascular changes of the choroid related to the elevation of intra-ocular pressure.

computer science, interdisciplinary applications,mathematical & computational biology,medical informatics, information systems

What problem does this paper attempt to address?

This paper aims to address the issues of automatic segmentation and visualization of the choroid in Optical Coherence Tomography (OCT) images. Specifically, the paper tackles the following two main problems:

1. **Difficulty in segmenting the choroidal-scleral interface (CSI)**: The CSI boundary in OCT images is often blurred, leading to inaccurate automatic segmentation.
2. **Interference from choroidal vessel shadows**: Projections of superficial retinal vessels cast shadows on the choroid, affecting its visualization.

To solve these problems, the authors propose a knowledge-infused deep learning method. This method includes the following aspects:

- **Biomarker-infused global-to-local network (Bio-Net)**: Used for choroid segmentation, it improves segmentation accuracy by using the predicted choroidal thickness as a constraint and employs a global-to-local strategy to provide overall structural information and suppress overfitting.
- **Deep learning pipeline for removing vessel shadows**: It first locates shadow positions using projections on the retinal pigment epithelium (RPE) layer, then uses an edge-to-texture generative adversarial repair network (Deshadow-Net) to predict the choroidal vessel content at the shadow locations.

Experimental results show that this method outperforms existing methods in both choroid segmentation and shadow removal tasks. Additionally, it demonstrates the ability to detect choroidal structural and vascular changes associated with elevated intraocular pressure in a clinical prospective study.