Automatic three-dimensional segmentation of endoscopic airway OCT images

Li Qi,Xipan Li,Wufan Chen,Zhongping Chen,Kaibin Zheng,Qianjin Feng

doi:10.1364/boe.10.000642

Abstract

Automatic delineation and segmentation of airway structures from endoscopic optical coherence tomography (OCT) images improve image analysis efficiency and thus has been of particular interest. Conventional two-dimensional automatic segmentation methods, such as the dynamic programming approach, ensures the edge-continuity in the xz-direction (intra-B-scan), but fails to preserve the surface-continuity when concerning the y-direction (inter-B-scan). To solve this, we present a novel automatic three-dimensional (3D) airway segmentation strategy. Our segmentation scheme includes an artifact-oriented pre-processing pipeline and a modified 3D optimal graph search algorithm incorporating adaptive tissue-curvature adjustment. The proposed algorithm is tested on endoscopic airway OCT image data sets acquired by different swept-source OCT platforms, and on different animal and human models. With our method, the results show continuous surface segmentation performance, which is both robust and accurate.

Highlights

Endoscopic optical coherence tomography is a high-resolution, label-free, and cross-sectional imaging modality that has already been used for disease screening of coronary arteries [1,2,3], gastrointestinal tract [4,5], and airway [6,7,8]
We presented an automatic three-dimensional airway segmentation scheme based on the 3D optimal graph search algorithm
The major contributions of our proposed processing scheme are: 1) comparing to 2D segmentation methods such as the dynamic programming algorithm, our approach could preserve the 3D continuity of the airway tissue surfaces; 2) we proposed an modified version of the 3D optimal graph search algorithm that utilized the slope of the airway as a constraint to achieve better surface detection results; 3) by performing the artifact-oriented pre-processing, the algorithm deals with various artifacts in the images, including reliable sheath removal and noise cancelling; and 4) special measures that dramatically cuts down the scale of the graph search problem are taken to reduce the computation cost, while the image resolution and surface detection accuracy are well preserved

Summary

Introduction

Endoscopic optical coherence tomography is a high-resolution, label-free, and cross-sectional imaging modality that has already been used for disease screening of coronary arteries [1,2,3], gastrointestinal tract [4,5], and airway [6,7,8]. The proposed scheme features an artifact-oriented pre-processing step, which removes the imaging artifacts such as the speckle noise and the plastic sheath, and at the same time obtains the surface slope of the airway lumen. This slope information is further used as a constraint for the subsequent graph search problem. The airway lumen is first detected, and used as a reference to flatten the structures under the airway surface This flattening procedure will dramatically reduce the size of the graph search problems for segmenting the mucosa and submucosa layers, and will be able to largely cut down the total processing time. The segmentation results can be 3D reconstructed for visualization and analysis

Image pre-processing The purposes of this step are

Multiple surface detection

Endoscopic airway OCT systems and OCT image data sets

Computational performance

Findings

Discussion and conclusion