Automated segmentation and quantification of airway mucus with endobronchial optical coherence tomography.

David C Adams,Hamid Pahlevaninezhad,Richard C Boucher,Josalyn L Cho,Andrew D Luster,Daniel L Hamilos,Melissa J Suter,Mehmet Kesimer,Margit V Szabari,Benjamin D Medoff

doi:10.1364/boe.8.004729

Abstract

We propose a novel suite of algorithms for automatically segmenting the airway lumen and mucus in endobronchial optical coherence tomography (OCT) data sets, as well as a novel approach for quantifying the contents of the mucus. Mucus and lumen were segmented using a robust, multi-stage algorithm that requires only minimal input regarding sheath geometry. The algorithm performance was highly accurate in a wide range of airway and noise conditions. Mucus was classified using mean backscattering intensity and grey level co-occurrence matrix (GLCM) statistics. We evaluated our techniques in vivo in asthmatic and non-asthmatic volunteers.

Highlights

Optical Coherence Tomography (OCT) is an interferometric imaging modality that can be thought of as the optical analog to ultrasound [1, 2]
To evaluate how the automated segmentation algorithm compared to manual segmentation, two expert OCT image readers were assigned to perform manual segmentation of a subset of 100 images from the human in vivo study
Images were assigned for manual segmentation by randomly selecting 5 asthmatic and 5 non-asthmatic data sets, and taking 10 cross-sectional images from each data set at fixed intervals along the pullback

Summary

Introduction

Optical Coherence Tomography (OCT) is an interferometric imaging modality that can be thought of as the optical analog to ultrasound [1, 2]. One important aspect with the potential for improving the clinical viability of the technology is the development of robust automated segmentation and analysis algorithms to speed data analysis and alleviate the burden of image interpretation. Such algorithms are routinely used in clinically established imaging technology such as CT, MRI, and ultrasound [11]. To our knowledge, has not been explicitly addressed in previously proposed algorithms, is the impact of the presence of fluid and particulate matter within the lumen on algorithm performance These factors are important to consider when measuring the physical dimensions of the system due to their influence on optical path length [17]. This becomes even more relevant if the organ system is diseased, as for example with asthma or chronic obstructive pulmonary disease (COPD) in the lung, where mucus production has been reported to correlate with disease severity [18]

Methods

Results

Discussion

Conclusion