Using Reactive Machine AI and Dynamic Changes in Lung Structure to Derive Functional Quantitative Lung CT Metrics of COPD, ILD, and Asthma


The previous chapter reviewed the lung CT AI methods to assess normal and diseased lung structure using a single TLC chest CT scan. COPD, ILD, and asthma can all produce small and large airway disease. This chapter will look at how lung CT AI can be used to assess lung ventilation by obtaining two chest CT scans with each scan taken at a different lung volume. This approach can be viewed as a dynamic assessment of the lungs that can provide information on lung function by assessing lung ventilation. Functional lung CT imaging can provide indirect information on the function of small airways, <2 mm in diameter, that cannot be measured directly by currently available CT scanners. Direct imaging of the airway tree can provide structural information on normal and diseased airway tree geometry (e.g., lumen area, wall area) at different generations from the trachea, generation 1, to subsegmental airways out to airway generation 6. Direct imaging of the airway tree by chest CT will be discussed later in this chapter.

Expiratory QCT Assessment of Air Trapping Due to Small Airway Disease in the Lung

Obtaining a chest CT scan at a lower lung volume, FRC or RV, can be used to assess air trapping in the lung, areas of the lung where air cannot freely be exhaled. The process of exhaling air out of your lungs decreases the volume of your lung by decreasing the amount of air per unit volume and since the tissue per unit volume is approximately constant, the overall density of the lung tissue normally increases when you exhale ( Fig. 6.1 ). The normal expiratory CT scan obtained at either FRC or RV shows increased density throughout both lungs due to the decrease in air volume in the lung without substantially changing the tissue volume. The lung decreases in size and the density of the lung increases to the extent that air is exhaled from each region of the lungs. This decrease in air is greater in the lower lobes than the upper lobes and is greater in the dependent portions of the lungs compared to the nondependent portions of the lungs ( Fig. 6.1 ).

Fig. 6.1

(A) Normal supine inspiratory (TLC) chest CT axial image. ( B) Normal supine expiratory (RV) chest CT axial. Note the increased density in the lungs in B due to the decreased air per unit volume of lung. The increase in lung density in B is greater in the more dependent portion of both the lungs (arrows) . Note the cross-sectional area of both lungs has also decreased in B .

(Courtesy of VIDA.)

When there is a narrowing of the small airways in the lung or a reduced number of small airways in the lung or both, the lungs will not exhale as much air as they normally would in the regions of the lung where there are abnormal small airways. The areas of lung where there is incomplete emptying of the air will decrease the density of the lung in those regions, compared to the adjacent normal lung tissue, on expiratory chest CT scans. This is referred to as air trapping and is seen on FRC/RV chest CT scans as regions of relative decreased density compared to normal areas of higher density on the expiratory CT scan ( Fig. 6.2 ). The trapped air can be quantitated in several ways and at different scales, including assessing both lungs together, individual lungs, lung lobes, and at the level of the individual lung voxel. Assessing air trapping at progressively smaller scales can increase spatial information that is lost when the air-trapping analysis is done by assessing both lungs together.

Fig. 6.2

Supine inspiratory (TLC) chest CT axial image (A) and supine expiratory (RV) chest CT axial image (B) from a patient with severe COPD (GOLD Stage 4). There is little change in lung density throughout both lungs in B compared to A (arrows in A and B ) . This lack of change in lung density is due to air trapping from obstruction and/or loss of small airways, <2 mm in diameter ( see text ).

(Courtesy of VIDA.)

The mean lung density on expiration, FRC/RV, CT scans can be used to assess air trapping. Similar to the LAA and HAA measures described in Chapter 5 , the amount of lung that is lower than a defined threshold HU value on an FRC/RV lung CT can assess the amount of air trapping in the lung ( Fig. 6.3 ). The ratio of the mean lung density on expiration to the mean lung density on inspiration can also be used to assess the amount of air trapping present in the lung. Air trapping can also be assessed in the lung by taking the ratio of the CT lung volume at FRC to the CT lung volume at TLC. A much more powerful way to assess the air trapping in the lung is to use nonrigid registration methods to compare voxel by voxel the differences in lung attenuation between the TLC and RV chest CT scans. Two well-known published methods that use image registration to assess air trapping at the voxel level: the parametric response map (PRM) and the disease probability map (DPM). The different methods to determine air trapping in the lung are summarized in Box 6.1 .

Fig. 6.3

RV lung CT voxel histogram curve taken from a normal patient (A) and a patient with air trapping from COPD (B) . The vertical red lines mark the −856 HU value ( A , LAA −856 = 0.1%, mean lung density (MLD) = −531 HU. B , LAA −856 = 66.7%, MLD = −854 HU).

Box 6.1

Qualitative CT Methods to Assess Air Trapping

  • 1.

    Mean value of the lung CT voxels on an FRC lung CT

    • This is referred to as the expiratory mean lung density (MLD expiration )

  • 2.

    Determine the percentage of FRC or RV lung CT voxels that are <−856 HU

    • This is referred to as the LAA −856

  • 3.

    Ratio of the FRC lung CT MLD to the TLC lung CT MLD

    • MLD expiration /MLD inspiration

  • 4.

    Ratio of the FRC CT lung volume to the TLC CT lung volume

    • CT FRC /CT TLC

  • 5.

    Determine the difference between TLC CT lung volume and RV CT lung volume

    • CT TLC -CT RV

  • 6.

    Functional small airway disease using PRM or DPM

    • fSAD

Whole Lung Assessment of Air Trapping Using LAA in Severe Asthma Patients

Air trapping is a hallmark of patients with asthma. Busacker et al. in 2009 reported the results of assessing air trapping in severe asthma subjects enrolled in the Severe Asthma Research Project (SARP) using quantitative assessment of chest CT scans obtained at FRC. Air trapping was defined to be significant in this study if 9.66% or more of the lung tissue was <−850 HU, LAA −850 , on the expiratory chest CT scans obtained at FRC. The quantitative chest CT assessment of LAA −850 in this study identified severe asthma subjects that were at increased risk for asthma-related hospitalizations, ICU visits, and mechanical ventilation. Multivariate analysis showed that those subjects with an LAA −850 >9.66% had an increased risk of asthma, pneumonia, high levels of airway neutrophils, and airflow obstruction measured by FEV1/FVC ratio and atopy.

Whole Lung Assessment of Air Trapping Using LAA in COPD Patients

Schroeder et al. in 2013 reported the results of doing TLC and FRC quantitative CT scans to assess air trapping and emphysema in 4062 COPD subjects enrolled in the COPDGene research study. The quantitative CT metric of air trapping was defined as the percent or fraction of lung tissue <−856 HU, LAA −856 , on the expiratory FRC CT scan ( Fig. 6.3 ). The results of this study showed that the LAA −856 metric of air trapping strongly correlated with decreases in airflow measured by spirometry, FEV1 R = −0.77 and FEV1/FVC R = −0.84, in this large group of COPD subjects. The corresponding LAA −950 metric of emphysema derived from the TLC chest CT scans (see Chapter 5 ) did not correlate as strongly with decreases in airflow, FEV1 R = −0.67 and FEV1/FVC R = −0.76.

Whole Lung Assessment of Air Trapping in the COPDGene 2019 Classes of COPD

As we discussed in Chapter 5 , the COPDGene 2019 Classes of COPD are determined by first defining four novel criteria to diagnose COPD: “Exposure, Symptoms, CT Structural Abnormality, Spirometry”. The CT structural criteria include having one or more of the following lung CT AI QCT metrics: LAA −950 equal to 5% or greater on a TLC chest CT scan (QCT structural measure of Emphysema), LAA −856 ≥15% on expiratory CT scan (QCT functional measure of air trapping), Pi10 equal to 2.5 mm or greater (QCT structural measure of airway wall thickening). The presence of 15% or more of air trapping on an expiratory FRC CT using the QCT metric LAA−856 in a patient with a 10-pack-year or greater smoking history, means they have at least possible COPD using the COPDGene 2019 Classes of COPD (see Chapter 5 ).

Whole Lung Assessment of Air Trapping Using MLD and CT Determined Lung Volumes

A recent publication by Pompe et al. reported assessing air trapping in 5697 COPD subjects enrolled in the COPDGene study using the whole-lung mean lung density (MLD FRC ) on FRC CT scans and also the ratio of the CT determined lung volume on FRC CT scanning to the CT determined lung volume on the TLC CT (CT-FRC vol /CT-TLC vol ). This study showed significant increases in (CT-FRC vol /CT-TLC vol ) and significant decreases in MLD frc , both indicators of whole lung air trapping, between baseline TLC and FRC CT scans and 5-year follow-up TLC and FRC CT scans in patients with COPD, GOLD Status 1–4. These results were attenuated but did not resolve when FEV1 was included in the statistical models. The authors indicate that the CT measures of air trapping in patients with COPD can progress without significant change in FEV1. CT determined air trapping are not completely determined by measures of FEV1 and the CT measures of air trapping in this study provide additional information compared to FEV1. The chest CT determined MLD FRC and CT-FRC vol /CT-TLC vol are both effective means of assessing air trapping and following the progression of air trapping over time in subjects with COPD.

Whole Lung Assessment of Air Trapping in Bronchiolitis Obliterans

Diseases of the small airways, <2 mm in diameter, are referred to as bronchiolitis. there are two predominate forms of bronchiolitis: cellular/proliferative bronchiolitis and obliterative/constrictive bronchiolitis. The primary causes of cellular bronchiolitis are infection, smoking, hypersensitivity pneumonitis, follicular bronchiolitis, and diffuse panbronchiolitis. The primary causes of bronchiolitis obliterans are postinfection, lung and bone marrow transplantation, connective tissue disease, toxic fume inhalation, adverse drug reactions, and inflammatory bowel disease. Bronchiolitis produces air trapping in the lung, which can be assessed with visual lung CT and lung CT AI methods.

The development of small airway disease in lung transplant recipients has been assessed using lung CT AI. The development of rejection in the transplanted lung can produce narrowing and obstruction of the lumen of small airways. This produces air trapping on expiratory CT scans that can be assessed using QCT metrics of air trapping. The development of small airway disease in lung transplants due to allograft rejection is described as bronchiolitis obliterans syndrome or BOS. Barbosa et al. reported in 2018 their retrospective experience in assessing 178 lung transplant patients who had 3D whole lung TLC and CT scans along with spirometry to assess FEV1, FVC, and FEV1/FVC ratio. Of these patients, 99 had a clinical diagnosis of BOS, and 79 were BOS negative.

The lung CT AI method in this study acquired two high-quality 3D chest CT scans with 1-mm slice thickness obtained at TLC and RV. The lungs on the TLC and RV CT scans were automatically segmented from the rest of the thoracic anatomy. The following quantitative CT (QCT) metrics were automatically generated for both lungs, individual lungs, and lung lobes: LAA -950 on TLC CT scans, LAA -856 on RV CT scans, and lung volume difference between TLC and RV CT scans. Nonrigid registration methods were used to identify the difference in attenuation values (HU) between corresponding voxels at TLC and RV. The number or volume of coregistered voxels that had changed by <75 HU were considered the optimal image registration-driven method for assessing air trapping from small airway disease. Whole-lung volume difference (WLVD), left-lung volume difference (LLVD), and right-lung volume difference (RLVD) had the strongest correlations with FEV1.

The study computed the multivariate Pearson correlations of multiple QCT metrics, as well as multiple visual metrics of BOS on the CT studies. This study showed quantitative CT determined whole lung volume difference (QCT WLVD) between the inspiratory and expiratory CT scans had the strongest correlation with FEV1, in patients who underwent bilateral lung transplants, R = 0.69 and P <0.0001. The QCT WVLD also had the strongest correlation with FEV1 in patients who had undergone a right lung transplant, R = 0.69 and P <0.0001, though QCT RLVD was not surprisingly very similar, R – 0.68 and P <0.0001. The QCT LLVD had the highest correlation with FEV1 in patients who underwent a left lung transplant, R = 0.86 and P <0.0001.

Assessment of Air Trapping at the Voxel Level Using Image Registration

It has been shown that the narrowing and loss of small airways of <2 mm in diameter in COPD subjects is the major site of obstruction to airflow. This loss of small airways is the major cause of air trapping assessed on expiratory chest CT scans. It has also been shown that the loss of terminal bronchioles precedes the development of emphysema in COPD patients. The assessment of air trapping using whole-lung, lung, and lobe expiratory CT scans obtained at FRC or RV in patients who also have emphysema, as is often the case in COPD patients, is an issue, since the areas of emphysema will be included in the air trapping index. The desire to detect air trapping separate from emphysema is to improve the detection of COPD-related lung injury before emphysema develops and to assess treatment effects on air trapping separate from emphysema. This has led to the development of other QCT metrics to assess air trapping apart from whole lung, lung, and lobe assessments. Box 6.2 summarizes the different scales that can be used to assess air trapping.

Feb 6, 2023 | Posted by in CARDIOLOGY | Comments Off on Using Reactive Machine AI and Dynamic Changes in Lung Structure to Derive Functional Quantitative Lung CT Metrics of COPD, ILD, and Asthma

Full access? Get Clinical Tree

Get Clinical Tree app for offline access