MSc Defence: The Effect of Anatomic Noise Associated with Ribs in Digital Chest Tomosynthesis

Student: Joshua Valerio

Supervisor: Dr. Jesse Tanguay

Abstract

Cystic Fibrosis (CF) is responsible for 90% of deaths from pulmonary failure, and may occur as early as in childhood, impacting quality of life. In tandem with current methods to diagnose CF patients, computed tomography (CT) is employed for thoracic imaging to monitor severity of lung symptoms, but its dose imposed on children is considered high risk and therefore not suitable. In this study, we propose using digital chest tomosynthesis (DCT) as an alternative to image CF patients as it imposes at most tenth of CT’s radiation dosage. However, due to DCT’s limited-angle CT scan methods, image quality is compromised due to its poor depth resolution and noise caused by out-of-plane artifacts. The rib cage bones in DCT images act as an overlying structure, which contribute a large amount of noise by blurring into the imaging plane as an out-of-plane artifacts, classified as anatomic noise. This impedes image quality, creating difficulties in monitoring the build-up of mucus as well as thickened airways. This dissertation focuses on quantifying this effect of anatomic noise using a 3-D printed rib cage phantom of a 15.5-year old male imported from the XCAT pediatric library.

The objectives of this study are to quantify the magnitude of noise using 3-D imaging by DCT and the frequency content of bone obtained through analysis of 2-D noise power spectrum (NPS) curves. We also show process of constructing the 3-D printed phantom as well as validating its material to closely reflect the attenuation properties of rib bone.

We quantified the magnitude of noise by varying the angular range using a fixed angular separation and conversely varying the angular separations within a fixed angular range. We evaluated the noise drop-off by measuring the blurring distance at half the noise magnitude as well as measuring the average max noise drop-off in the lung region. It was determined that decreasing the angular range had a larger increase in noise blur in the depth direction than varying the angular separation with values as high as 36.7%. We also reported κ values in the order of 10-7mm2and β-values averaging ~4.2 across five tube voltages. These results may help develop bone suppressing algorithms to better the process of tomosynthesis for imaging CF pediatric patients.

Keywords: 3-D printed phantom, noise drop-off, anatomic NPS, cystic fibrosis, pediatrics