Relationship of Cardiac Function with Pulmonary Function and Lung Volume using Whole-body MR Imaging: A Population-based Study

To evaluate the relationship of cardiac function, including time-volume-curves, with lung volumes derived from pulmonary function tests (PFT) and MRI in subjects without cardiovascular diseases. Methods 216 subjects underwent whole-body MRI and spirometry as part of the KORA-FF4 cohort study. Lung volumes derived semi-automatically using an in-house algorithm. Forced expiratory volume in one second (FEV1), forced vital capacity (FVC), and residual volume were measured. Cardiac parameters derived from Cine-SSFP-sequence using cvi42, while left ventricle (LV) time-volume-curves were evaluated using pyHeart. Linear regression analyses assessed the relationships of cardiac parameters with PFT and MRI-based lung volumes.


Background
To evaluate the relationship of cardiac function, including time-volume-curves, with lung volumes derived from pulmonary function tests (PFT) and MRI in subjects without cardiovascular diseases.
Methods 216 subjects underwent whole-body MRI and spirometry as part of the KORA-FF4 cohort study. Lung volumes derived semi-automatically using an in-house algorithm. Forced expiratory volume in one second (FEV1), forced vital capacity (FVC), and residual volume were measured. Cardiac parameters derived from Cine-SSFP-sequence using cvi42, while left ventricle (LV) time-volume-curves were evaluated using pyHeart. Linear regression analyses assessed the relationships of cardiac parameters with PFT and MRIbased lung volumes.

Conclusion
Subclinical cardiac impairment was associated with reduced FEV1, FVC, and residual volume. Cardiac parameters decreased with increasing MRI-based lung volume contrasting the results of PFT.

Background
Chronic respiratory diseases, such as chronic obstructive pulmonary disease (COPD) including emphysema, are among the leading causes of morbidity and mortality [1], claiming some 2.5 million lives worldwide in 2015 [2]. Moreover, chronic respiratory diseases increase the cardiac overload and may result in cardiac impairment, thus increasing morbidity and further affecting quality of life [3]. Therapeutic options for chronic respiratory disease and consequent cardiac impairment are limited and increase health care costs. Therefore, early detection and management are important measures to slow COPD progression, exacerbations, and hospitalizations [4].
Recent studies reported on the association of impaired cardiac function and maladaptive deformation with respiratory parameters. Watz et al. showed decreased cardiac chamber sizes and impaired left ventricular diastolic lling patterns as well as global right-heart dysfunction in patients with known hyperin ation [5]. Another study of cardiac volumes assessed by magnetic resonance imaging (MRI) in patients with severe emphysema showed reduced left and right ventricle end-diastolic volumes, suggesting an association with hyperin ation [6]. Larger cohorts such as the Multi-Ethnic Study of Atherosclerosis (MESA), which used computed tomography (CT) for assessment of the lung volume as well as MR imaging for cardiac assessment, further characterized the relationship between cardiac impairment and COPD. In this study within MESA, Barr et al. observed that emphysema severity was linearly related to impaired left ventricular lling, reduced stroke volume, and lower cardiac output, but without change in the ejection fraction [7].
While most studies refer to diastolic dysfunction assessed by increased left-ventricular (LV) end-diastolic volume, the right ventricular (RV) function was assessed in the same MESA study. Pulmonary hyperin ation was associated with smaller RV end-diastolic volume, stroke volume, cardiac output, and reduced RV mass. An increase in RV afterload was observed among current smokers [8], potentially related to diminished pulmonary vascular bed due to apoptosis of pulmonary endothelium [9], or endothelial dysfunction [10], as well as an increase in pulmonary vascular resistance, among others. However, cardiac time-volume-curves have so far not been assessed in this context. The diastolic function can be more precisely characterized by measuring early and late ventricular diastolic lling rates. Cardiovascular magnetic imaging can result in a separate volume calcualtion at each phase, and the lling rate curve (dV/dT) at different parts of the cycle can evaluate diastolic function [11]. Detection of subclinical cardiac impairment, especially in the early stages of COPD, is key to reduce disease burden. Melerba et al. reported LV diastolic dysfunction using Doppler-echocardiography in patients with early-stage COPD and with no clinical signs of cardiovascular dysfunction [12]. Moreover, Thomson et al. reported that reduced forced expiratory volume in one second (FEV1) and forced vital capacity (FVC) by spirometry were associated with smaller ventricular volumes and reduced ventricular mass determined by cardiovascular MRI [13].
While pulmonary functional tests (PFT) are widely available for evaluating respiratory function, advanced imaging by radiation-free visualization techniques, such as whole-body MRI are also now accessible. For research purposes, whole-body MRI may represent a promising imaging technique that enables cardiac and pulmonary assessment at the same time [14]. Recently, our KORA (Cooperative Health Research in the Region of Augsburg) cohort study suggested that lung volume derived from whole-body MRI is associated with PFT-derived residual volume, but also with the FEV1/FVC ratio (Tiffeneau index), as well as with a clinical history of COPD [15]. Based on this evidence, we explored the clinical value of MRI-derived lung volumes based on a short, non-dedicated T1-Dixon sequence, within the framework of the KORA MRI study, which excluded subjects with known cardiovascular disease. As the rst study to evaluate cardiac parameters and lung volumes from one single whole-body MRI scan, we aimed rstly to assess the association of MRI-based cardiac parameters with FEV1, FVC, and residual volume as derived from PFT, and secondly to evaluate the relationship between these cardiac parameters and MRI-derived lung volume.

Study Population
The KORA FF4 cohort study represents a broad sample from a general population in the region of Augsburg, Germany. The study recruited subjects (n = 1,851) aged between 25-74 years, and participants were examined between June 2013 and September 2014 at the KORA study center [16]. An MRI scan (3 T) was incorporated in a sub-study within 400 subjects who underwent whole-body MRI. Exclusion criteria were contraindications to MRI, gadolinium administration, or a known cardiovascular disease. Due to incomplete MRI data and/or inadequate image quality, 22 subjects were excluded [17]. In the current analyses, 378 subjects were included.

Pulmonary Function Test
Pulmonary function tests were performed in line with the American Thoracic Society and European Respiratory Society recommendations [19,20]. Flow-volume curves were acquired using a pneumotachograph-type spirometer (MasterScope, Jaeger, Hoechberg, Germany). Subjects (n = 216) performed at least 3 and up to 8 spirometry maneuvers to obtain a minimum of two acceptable and reproducible parameters. Residual volume was determined within a measurement of lung diffusing capacity using the single-breath technique with subjects performing a maximum of ve trials in order to achieve a minimum of two acceptable and reproducible maneuvers with an effective breath-hold time within 10 ± 2 s.

Whole-Body MR Imaging
Whole-body MRI scans were performed with a 3-Tesla MRI system (Magnetom Skyra, Siemens AG, Healthcare Sector, Erlangen, Germany) [17]. The protocol comprised sequences covering the entire body (from neck to below hip) for tissue/organ evaluation. MR Image Analysis for Lung Volume As described previously [15,21], an algorithm was used for the automatic procession of the MR data.
Brie y, the axial in-phase sequence provided the basis to process lung segmentation including important aspects to ensure high-quality outputs, such as pre-extraction of a coarse region of interest comprising the airways, adjustment of general intensity inhomogeneities, smoothing of the lung and morphological operations, and last but not least 3D Vesselness Filter-enhanced trachea extraction. For further quality assurance, four independent experienced radiologists manually read seventeen randomly selected data sets, while each data set was evaluated by two different radiologists. Using this self-developed algorithm mentioned above, the readers were unaware of the clinical covariates and performed analyses in a blinded fashion at all times.

MR-Image Analysis of Cardiac Measurements
LV and RV function were evaluated using commercially available software (cvi42; Circle Cardiovascular Imaging, Calgary, Alberta, Canada) by two independent readers. After manually segmenting the lumens for RV end-systoles and end-diastole in each layer, the software calculated automatically the corresponding volumes. The difference between the end-systolic and end-diastolic volumes comprises the parameters for stroke volume and ejection fraction. Detection of LV contours and calculation of LV volumes was processed automatically, and if necessary, corrected manually according to current guidelines [22]. LV myocardial mass was assessed during end-diastole. Normal values were referenced from a recent publication [23].
Furthermore, lling and ejection rates for LV were quanti ed using pyHeart, dedicated in-house software displaying LV time-volume-curves (Fig. 1). Peak gradients were anticipated during systolic ejection as well as early LV lling, which is mainly a passive process, and late LV lling, which is driven by atrial contraction [11].

Statistical Analysis
Clinical characteristics and MRI parameters of cardiac and lung structure and function were presented as arithmetic means with standard deviation (SD) for continuous variables or as counts and percentages for categorical variables. We compared the characteristic differences among tertials of the lung volumes by one-way ANOVA or chi 2 -test. Using linear regression models, cardiac MRI parameters as exposure variables were assessed in relation to FEV1, FVC, residual volume and MRI-derived total lung volume as outcomes.
First, in the adjusted model we included BSA. Second, the fully adjusted model included adjustments for age, sex, BSA, and smoking status, providing β-coe cients with 95% con dence intervals (CI). A p-value of less than 0.05 was considered statistically signi cant with regard to all analyses. Statistical analyses were performed using the Stata 14.1 software package (Stata Corporation, College Station, TX, U.S.A.). Table 1 presents the clinical characteristics of the study population with a mean age of 56.3 ± 9.2 years and 57% male subjects. 44% of subjects were former smokers, 20% were current smokers, and 36% never smokers. Based on PFT, the average residual volume was 2.2 ± 0.4 L while FEV1 was 3.1 ± 0.8 L, and FVC was 4.2 ± 1.0 L ( Table 1). Total lung volume derived from whole-body MRI was 4.0 ± 1.1 L; overall the right lung volume was larger than the left lung volume at 2.2 ± 0.6 L and 1.8 ± 0.5 L, respectively (p < 0.001). Cardiac parameters were within expected ranges (Table 1). Only 9 subjects had an abnormally increased LV enddiastolic volume. Regarding the RV, 17 subjects had an abnormally increased end-diastolic volume; there was some overlap between LV and RV increased end-diastolic volume (n = 6). Based on time-volume curves ( Fig. 1), the LV peak ejection rate was 354.9 ± 133 mL/s, while the early and late diastolic lling rates were on average similar (226 ± 115 and 237 ± 140 mL/s, respectively; Table 1). The correlation between early and late diastolic lling rate was 0.25 (p < 0.001) and both parameters correlated with increased LV enddiastolic volume (r = 0.68, p < 0.001; r = 0.29, p < 0.001, respectively), and also with RV end-diastolic volume (r = 0.60, p < 0.001; r = 0.17, p < 0.001, respectively). No reference values were available for the assessment of the time-volume curves.

Association between Cardiac Parameters and Pulmonary Function
In the BSA-adjusted model, LV cardiac parameters were associated with FEV1, except for ejection fraction, which was inversely associated, and late diastolic lling rate (non-signi cant). In a fully adjusted model, cardiac parameters remained associated, except for myocardial mass, ejection fraction and late diastolic lling rate (Table 2). Similarly, in the BSA-adjusted model RV cardiac volumes were associated with FEV1, and after full adjustment only ejection fraction became non-signi cant. Figure 2 shows the association of LV end-diastolic volume and LV early diastolic lling rate with FEV1 in an unadjusted model. Further, when assessing the relationship between LV cardiac parameters and FVC in the BSA-adjusted model, all parameters were associated with FVC, except for ejection fraction, which was inversely associated, and late diastolic rate (non-signi cant). In the fully adjusted model, the same parameters remained associated, excluding myocardial mass and ejection fraction, but the association with late diastolic lling rate became signi cant. Furthermore, RV cardiac parameters were associated with FVC. But after fully-adjustment, ejection fraction became non-signi cant (Table 2). Moreover, LV end-diastolic volume, end-systolic volume, ejection fraction, peak ejection rate, and early diastolic lling rate were associated with residual volume in the BSA-adjusted model, while again ejection fraction was inversely associated. In a fully adjusted model end-diastolic volume, peak ejection rate, and early diastolic lling rate remained associated, while the late diastolic lling rate turned to be associated as well. While in the BSA-adjusted model RV end-diastolic volume and end-systolic volume were associated with residual volume, after further adjustment most RV parameters were no longer associated, except for stroke volume becoming signi cantly associated ( Table 2). No associations were seen with the Tiffeneau index (data not shown). The beta estimate given with a 95% con dence interval represents the association size between cardiac measurements and pulmonary function tests. The fully adjusted model includes the following covariates: age, sex, BSA, and smoking status. Abbreviation: BSA = Body Surface Area. CI = 95% con dence interval; FEV1 = forced expiratory volume in the rst second; FVC = forced vital capacity; SD = standard deviation. Bold font indicates statistical signi cance (P < 0.05)
When assessing the association of RV cardiac parameters with lung volume, in the BSA-adjusted model only ejection fraction was inversely associated. However, with the additional fully adjusted model, ejection fraction was no longer associated with lung volume, while stroke volume (β= − 0.11, p = 0.01) became inversely associated with lung volume (Table 3).  The beta estimate given with a 95% con dence interval represents the association size between cardiac measurements and MRI derived total lung volume. The fully-adjusted model includes the following covariates: age, sex, BSA, and smoking status. Abbreviation: BSA = body surface area. Bold font indicates statistical signi cance (P < 0.05).

Discussion
In this sample of subjects without clinical history of cardiovascular disease drawn from a general population, we found that several MRI-based cardiac parameters were associated with lung volumes derived from pulmonary function tests. Interestingly, MRI-based cardiac parameters were inversely associated with MRI-based lung volumes. We observed a notable relationship of FEV1, FVC, and residual volume determined by PFT as well as MRI-based lung volumes with selected cardiac measures.
Firstly, we observed an association between end-diastolic and end-systolic volumes, and stroke volume for both LV and RV with FEV1 and FVC, as well as with peak ejection rate, and early diastolic lling rate; additionally, late diastolic lling rate showed an association with FVC. No associations were seen between cardiac parameters and the Tiffeneau index. Secondly, there were signi cant relationships between LV enddiastolic volume, RV stroke volume, peak ejection rate, and early and late diastolic lling rate, with residual volume. For PFT, our results are consistent with the recent ndings from Thomson et al. who showed, using the UK Biobank data, that lower FEV1 and FVC were associated with smaller LV end-diastolic, end-systolic, and stroke volumes, as well as RV end-diastolic, end-systolic, and stroke volumes [13]. only affect end-diastolic volumes, but also the diastolic lling rates, which has not been reported so far. We also observed a correlation between the late diastolic lling rate and residual volume, suggesting the impact of residual volume on increased left atrial lling pressures and left atrial contractile function. To our knowledge, this study is the rst to describe an inverse association of lung volume with the early diastolic lling rate derived from whole-body MR scans.
With respect to the lung volume derived from whole-body MRI, stroke volume was inversely associated for both LV and RV. For the LV, end-diastolic volume was inversely related; the early, but not the late diastolic lling rate, was inversely related to MRI-based lung volume. Overall, we observed a signi cant inverse association between lung assessment by MRI and cardiac parameters, in contrast with lung assessment by PFT.
MRI-derived lung volume shows a good correlation with residual volume (r = 0.57) and is independently associated with obstructive ventilatory impairment, based on PFT measurements (Tiffeneau index) and clinical presentation [15]. Whereas the Tiffeneau index is the established parameter to de ne bronchial obstruction, residual volume can be an indicator of hyperin ation in obstructive lung diseases, and has a unique prognostic value in COPD patients [24].
However, the inverse association between cardiac parameters and MRI-based lung volume contrasted with the positive associations between cardiac parameters and PFT. More speci cally, in subjects with both LV end-diastolic volume and early LV diastolic lling decrease, we observed an increased MRI-based lung volume. Again, both cardiac parameters increase with increased residual volume (similarly for FEV1 and FVC Our ndings are supported by previous evidence, as an inverse association between clinically diagnosed severe emphysema and end-diastolic volume for LV measured by cine MRI was described earlier, despite the relatively small number of patients with emphysema (n = 24) [6]. Our results also showed early LV lling impairment and low stroke volumes for both LV and RV, with no change in ejection fraction, in association with lung volume. This is consistent with previous studies where lung volumes were determined by CT and related cine MRI parameters [7,8], or ECG-gated CT angiography [25]. Also, peak LV lling rates in early and late diastole can be derived from rates of change in chamber volume -a technique made possible by the high spatial resolution of cine MRI (Fig. 1). The peak early and late lling rate in LV are sensitive markers and can indicate early subclinical diastolic dysfunction [26]. However, measuring these volume-derived indices can be a time-consuming process even with advanced software, and therefore impractical in routine clinical care [27]. LV diastolic function indices derived from whole-body MRI may have a future role in screening for early subclinical diastolic, especially if coupled with machine learning techniques.
One strength of our study is the use of an advanced MRI technology 3-Tesla generation, which includes the most advanced imaging modality to-date with well-de ned imaging protocol, image processing, and detailed information on the health condition of the study population. Previous studies quanti ed the lung volumes using CT, which contains radiation-exposure. The lung volume assessment using MRI provides an alternative radiation-free method for large-scale imaging studies. However, our study contains limitations which should be considered. Firstly, our study is a cross-sectional analysis design and therefore does not allow to assess relationships between alterations in pulmonary and cardiac parameters over time.
Secondly, although our sample size included about 400 subjects, this number is relatively small due to the laborious nature of image processing of whole-body MRI assessment. However, the ndings observed in our sample may represent important information for hypothesis generation. Thirdly, in our sample, only Caucasian participants with no history of cardiovascular disease were included, which may limit the generalizability of our ndings.

Conclusion
Subclinical cardiac impairment was associated with decreasing FEV1, FVC, and residual volume. Cardiac parameters decreased with increasing MRI-based lung volume contrasting the results of PFTs. This suggests that MRI-based lung volume may provide additional information on pathophysiological characteristics beyond PFTs.
Declarations Figure 1 Early and late diastolic lling rate of the left ventricle (slope) In ECG-triggered cine MRI, the lling rates were measured during diastole and assessed by pyHeart, displaying left ventricle (LV) volume versus time curve (middlebox). Example for segmentation of LV in 2-chamber short-axis views above. * = LV end-diastolic volume, # = LV end-systolic volume, yellow line = early diastolic lling rate, blue line = late diastolic lling rate.