Electronic Thesis and Dissertation Repository

Thesis Format

Monograph

Degree

Master of Engineering Science

Program

Biomedical Engineering

Supervisor

Ouriadov, Alexei

Abstract

Obstructive lung diseases are characterized by heterogenous ventilation. Hyperpolarized 129Xe gas lung magnetic resonance imaging (MRI) can examine lung ventilation heterogeneity by acquiring high-resolution isotropic images. The current gold standard of semi-automated (SA) segmentation can be used to quantify non-isotropic 129Xe lung images to generate ventilation defect percent (VDP), however, this method is not suitable for analysis of isotropic voxel 129Xe images due to the large number of slices. Therefore, we used a fully automated deep learning-based (DL) lung algorithm to calculate VDP from isotropic images. SNR, SA and DL-based VDP were calculated, showing a strong positive linear correlation with a zero intercept and close to unity slope. This study demonstrates the feasibility of using DL-based segmentation methods to quantify ventilation defects, which has potential for clinical translation of 129Xe MRI as a tool for treatment and monitoring for patients with pulmonary diseases.

Summary for Lay Audience

Obstructive lung diseases affect millions of individuals and include symptoms such as chronic cough, shortness of breath and frequent respiratory infections. There is currently no cure for many obstructive lung diseases, however patient treatment focuses on reducing symptoms and hospitalizations, as these diseases place a significant burden on healthcare across Canada. More recent lung diseases such as, COVID-19, directly affects the lungs, by damaging and destroying its cells. This is similar to other obstructive lung diseases, which result in lungs becoming inflamed and failure of gas exchange and respiratory function, which can lead to organ failure. Spirometry is widely available and is commonly used to diagnose obstructive lung disease, however, it only provides global lung function information.

Medical imaging techniques such as computed tomography (CT) and hyperpolarized gas magnetic resonance imaging (MRI) are used to study pulmonary diseases, as they provide regional lung information which spirometry cannot. Chest CT images can provide structural changes within the lungs, mainly in tissues and airways. Hyperpolarized gas MRI allows for visualization of lung structure as well as function, as it can detect unventilated regions of the lung, known as ventilation defects. Ventilation defects are quantified by the ventilation defect percent (VDP), which is calculated as the total ventilation defect volume (VDV) to the total thoracic cavity volume (TCV). Semi-automated (SA) segmentation methods are typically used for calculating VDP from lung images, however, this technique is difficult for analyzing VDP from isotropic high-resolution images as it is a time-consuming task. Recently, deep learning (DL) methods have demonstrated numerous successes in medical image analysis tasks. In this study, we acquired high-resolution 3D 129Xe data from participants with ventilation defects and calculated the VDP using a DL-based algorithm in comparison with a SA approach as the reference gold standard. We observed a strong linear correlation between the two types of VDP estimates. This study suggests that 129Xe MRI coupled with the DL-based lung-segmentation can be used to rapidly quantify ventilation defects across a wide range of disease. Medical imaging techniques such as computed tomography (CT) and hyperpolarized gas magnetic resonance imaging (MRI) are used to study pulmonary diseases, as they provide regional lung information which spirometry cannot. Chest CT images can provide structural changes within the lungs, mainly in tissues and airways. Hyperpolarized gas MRI allows for visualization of lung structure as well as function, as it can detect unventilated regions of the lung, known as ventilation defects. Ventilation defects are quantified by the ventilation defect percent (VDP), which is calculated as the total ventilation defect volume (VDV) to the total thoracic cavity volume (TCV). Semi-automated (SA) segmentation methods are typically used for calculating VDP from lung images, however, this technique is difficult for analyzing VDP from isotropic high-resolution images due as it will be a time-consuming task. Recently, deep learning (DL) methods have demonstrated numerous successes in medical image analysis tasks.

In this study, we acquired high-resolution 3D 129Xe data from participants with ventilation defects, using a previously developed key-hole-based method and calculated the VDP using a DL-based algorithm in comparison with a SA approach as the reference gold standard. We observed a strong linear correlation between the two types of VDP estimates. This study suggests that 129Xe MRI coupled with the DL-based lung-segmentation can be used to rapidly evaluate ventilation heterogeneity across a wide range of disease.

Download

Share

COinS