Image Quantification
Baseline HRCT scans of the patients and controls obtained between 2007
and 2019 were retrieved from the radiology archive. All CT scans were
performed non-contrast and volumetric thin-section using a 64-slice
multiple-detector CT scanner (Aquilion, Toshiba, Tokyo, Japan).
Participants were in the supine position and held a full inspiration
during acquisition. Images of 1 mm thickness were presented at window
settings designed for lung parenchyma evaluation (width 1,500 Hounsfield
units (HU); level – 500 HU). Scans were analyzed by Myrian XP Lung 3D
software (Intrasense SA, Montpellier, France) for lung volume and lobar
density quantification. One operator (OC) blinded to the patients’
clinical features performed lung volume and density quantification using
the Myrian toolbox. Myrian is a medical imaging and computer-aided
diagnostic software program used to view, store, reproduce, and export
medical images. It is used for 3D imaging involving maximum intensity
projection and extremely accurate volume mapping, which utilizes
dedicated segmentation algorithms to separate entire vascular pulmonary
structures and normal parenchyma and quantify high precision lung
parenchyma volume. The software is compliant with DICOM images and shows
the 3D images in three planes (axial, coronal, and sagittal). All CT
scans were pre-processed using Gaussian smoothing for noise reduction
and histogram equalization for contrast enhancement. CT data of the lung
was saved as DICOM format, and subsequently imported into Myrian-Xp-Live
software. The system automatically identified and extracted image
information of the airways, vessels and lung parenchyma, and then
generated reconstructed 3D images of the lung. The right and left lungs,
as well as the trachea and adjacent vasculature could be clearly
observed in all directions. Segmentation of lung volume with an adaptive
area was used to divide the left and right lungs. The option of manual
correction was used to solve problems caused by individual anatomical
variability at each segmentation step. Based on their high density, the
pulmonary vessels were extracted from the segmented lung volumes. For
the right and left lungs, the densities were measured automatically,
with a choice of – 700 or – 900 HU as the characteristic density value
of healthy lungs (Figure 1).