This study combines AI convolutional neural network techniques with measures of cardiac structure and function to demonstrate that the CT scan-measured subsegmental and segmental vasculature volume indexes correlate with echocardiographic measurements of cardiac structure and function in patients with COPD. This is the first study, to our knowledge, to specifically investigate this relationship in a cohort of well-characterized patients with COPD.

We found that patients with the lowest subsegmental vessel fraction i.e., more extensive pruning, had a higher extent of emphysema, had higher LVMI, due to thicker myocardial walls and smaller LV cavity size, higher IVC diameter, and higher LAVI, than patients with higher subsegmental vessel fraction. We found in the multivariable analysis that lower subsegmental vessel fraction was significantly linearly correlated with higher LV mass parameters, higher right atrial pressure, and impaired RV systolic function. In the multivariable analysis, we found segmental vessel fraction to be significantly correlated with higher LV mass parameters, higher right atrial pressure, and pulmonary pressure.

TAPSE was the only RV measure that showed a statistically significant correlation with subsegmental vessel fraction, suggesting that fewer subsegmental vessels with a CSA under 10 mm2 also have some consequence on the RV function and myocardial contractility. IVC diameter was found to be significantly negatively correlated with the subsegmental vessel fraction, suggesting that a lower amount of subsegmental vessels also could lead to a higher right atrial pressure.

The segmental vessel fraction was positively correlated with LVMI, LVPWD, and IVSd. Even after adjustment for clinical variables known to cause left-sided mass alteration, the measures remained significantly correlated, suggesting a higher blood volume in the larger vessels closer to the heart also impacts LV mass with a potential consequence of LV hypertrophy. TR max PG was the only RV measure significantly correlated with the segmental vessel fraction, suggesting that a lower number of segmental vessels could have a consequence of pulmonary vascular resistance and pulmonary pressure. Segmental vessel fraction was also significantly correlated with IVC diameter, suggesting that a lower number of segmental vessels also leads to a higher right atrial pressure.

The TPVV was positively associated with LVMI, LVPWD, IVSd and LVEDVi. However, after adjustments only LVPWD and LVEDVi remained significantly associated, suggesting that change in the total vascular volume in the lungs impacts both the LV mass and – volumes. Notably TPVV was the only vascular volume parameter associated with LV volume index, suggesting that the total vascular volume in the lung impacts the global LV diastolic performance. In general, oppositely directed correlations were found with subsegmental vessel fraction and segmental vessel fraction, which support the idea of large vessel dilatation as a compensatory mechanism to pruning of the small peripheral vessels. However, segmental vessel fraction was correlated to TR max PG while subsegmental vessel fraction was not. Subsegmental vessel fraction was correlated with TAPSE, while segmental vessel fraction was not. This suggests that the higher blood volume in the larger segmental vessels closer to the heart could have hemodynamic consequences on the RV and that segmental vessel fraction may be a more sensitive marker of higher pulmonary pressure. Furthermore, we found the subsegmental and segmental vessel fractions to be associated to with greater number of cardiac parameters compared to TPVV, which may suggest that the indexed vessel fractions are more sensitive markers of changes in cardiac structure and function.

Previous studies have primarily focused on small vessel fraction (small vessel volume/TPVV), regarding the severity of lung disease and less on cardiac impairment. Small vessels were defined as blood volume in vessels with nominal sizes below 5 mm2 in CSA. In these studies, a lower small vessel fraction was associated with the severity of airway disease, including impaired spirometry parameters, higher GOLD stage, higher risk of exacerbation, RV dysfunction, and higher all-cause mortality [2, 4, 7, 8].

A previous study by Wells et al. [2] investigated the association between small vessel fraction and cardiac function assessed by cardiac MRI in 24 patients with COPD. They found that small vessel fraction was significantly correlated with RV end-systolic volume index and RV mass index, and trending towards correlation with RVEF. We found TAPSE, a conventional echocardiographic surrogate of RVEF, to be significantly associated with the subsegmental vessel fraction. Both studies suggest that a lower number of small vessels in patients with COPD is associated with structural remodeling of the RV and impaired RV function. They did not find small vessel volume, nor the TPVV alone to be associated with any cardiac measures on MRI, suggesting that the indexed vessel fraction is a more sensitive marker of the lung-cardiac interplay than subsegmental vessel volume or TPVV alone. The vessel fraction being a more sensitive marker than the TPVV might be due to pruning causing a shift of blood from the smaller subsegmental vessel to the segmental vessels.

A study by Washko et al. [8] found similar results among previous and current smokers, where they found lower arterial volume to be independently associated with larger RV, with a 10 ml decrease in arterial small vessel volume giving a 1 ml increase in RV size measured with cardiac CT imaging. However, stratified multivariable models only found this to be true in patients with mild lung function impairment defined as an FEV1 above 73.6% of predicted. Furthermore, they found RV enlargement to be associated with a 63% higher risk of mortality, but only in patients with arterial pruning. No significant increased risk of death was seen without pruning [8]. Our study did not find any difference in FEV1 between the group of patients with the most extensive pruning and those with less extensive pruning. In addition, the GOLD stage did not correlate with the extent of pruning. This is probably due to the small number of patients in each GOLD stage group.

Another sub-study from the COPDgene study, by Rahaghi et al. [17] investigated the relations between small and large vessel fraction and RV function in 80 smokers using cardiac MRI. They found that both lower number of arterial and venous vessels were associated with dilation and hypertrophy of the RV with reductions in RV ejection fraction. They also found that an increase in arterial large volume fraction was associated with RV dysfunction. They suggested that the relationship between small vessel fraction and RV ejection fraction appeared to be driven by the arterial small vessel fraction [17]. They found that the arterial volume fraction remained of predictive power, after adjustment for the extent of emphysema. In our study, the extent of emphysema visible on CT scan as assessed by %LAA-950 was likewise minor, suggesting that microvascular destruction might be present before CT identified emphysema.

The existing literature suggests that LV systolic function is usually preserved in patients with COPD, whereas RV dysfunction is more common. However, like our study, a previous study [18] has found an association between pulmonary vessel volume and higher LV mass dimensions and lower LV filling. Aaron et al. [18] conducted a population-based study that investigated cardiac function with MRI and its association with TPVV. Reductions in total pulmonary vascular volume were associated with reduced left atrial volumes, reduced LV end-diastolic volume, and cardiac output, among previous and current smokers, including those without lung disease. They also found a lower TPVV to be associated with a greater LV mass/end-diastolic volume ratio, a parameter described as useful for the evaluation of LV concentric remodeling. Our study did not find any differences in the percentage of smokers (previous and current) in the group of patients with the most extensive pruning compared to those with less extensive pruning.

These consistent radiographic results likely indicate that pulmonary vascular injury which occurs as a consequence of the development of COPD affects cardiac structure and function. CT scans may hold the potential of providing us with more information about cardiovascular risk amongst COPD patients, hence, it might be possible in the future to deduce which patients are at higher risk for cardiac impairment and therefore must be followed more closely or started in prophylactic treatment. Our study suggests that patients with more extensive pruning and greater compensatory dilation have more pathological cardiac remodeling and may potentially gain from further examination and closer follow-up.

Current techniques and measurements are based on images that are already acquired during a standard initial workup of COPD patients, making them easily implemented in clinical practice. The subsegmental vessel fraction along with other risk markers may help guide which patients may benefit from additional imaging e.g., echocardiography. This could potentially identify the patients who require further examination and closer follow-up.

Major strengths of our investigation are the prospective inclusion of patients in the cohort and the well-described patient group. In addition, this study includes the use of novel quantitative pulmonary vascular measures.

Our study is limited by its relatively small sample size and single-center experience, though we did demonstrate correlations between pulmonary vascular injury and cardiac structure and function in accordance with existing literature. In addition, patients did not have the echocardiographic examination and CT scans within a set period of time. Changes in cardiac function and anatomic parameters between the CT and echocardiogram may occur.

A considerable number of patients (n = 23) were excluded from the analysis due to inadequate computational vascular reconstruction which introduces the risk of selection bias. Although the exclusion was merely due to technical factors, additional studies with larger sample sizes are needed to define the complex relationships more definitively between CT measurements and cardiac dysfunction in patients with COPD.