The true volume of in vivo pulmonary nodules is unknown, and even when those nodules are surgically excised, their volume changes because of the sudden stop in blood flow [17]. Therefore, it is our expectation that the blood flow inside and surrounding a given pulmonary nodule will be included in the segmentation result provided by the volumetry tool. In that sense, it is expected that an increase in the blood pressure felt inside the nodule will translate into an increase in the blood volume included in the segmentation.
Our study identified factors related to cardiopulmonary circulation, namely the location of the PN, the difference of the MPA diameter between systole and diastole, and vascular distance between the MPA and the PN, which were statistically associated with the measurements of volumetry tools.
These results could be explained by the propagation of vascular pressure from the heart to the nodule through the complex arterial and capillary network. Assuming the pressure difference at the MPA between systole and diastole, one might expect a similar pressure difference at the capillaries surrounding the PN and that the cardiac cycle phase during acquisition would also be important. However, our results do not support this intuition nor corroborate earlier results from Boll et al. [18]. This is only the second study investigating PN volumetry using ECG-gated scans to the best of our knowledge. The significant dependence of volume measurement on the cardiac cycle phase, reported previously, has not yet been independently validated. Differences between our study and the former include larger sample size, number of detector rows of the CT scanner (128 vs. 16 slices), and current and updated segmentation algorithms in clinical use. The independence of volume measurement from the cardiac cycle phase means that our data cannot be used to recommend any particular cardiac phase for PN volumetry on CCTA scans. It could also suggest that these hemodynamic effects could be seen in non-gated scans (such as LD-CT). The high temporal resolution of CCTA scans accounts for their sensibility to momentary cardiopulmonary hemodynamic changes (subtle differences between two phases in a single cardiac cycle). These momentary changes could be used to model clinically significant hemodynamic changes that a patient may experience when suffering from an acute or progressing cardiopulmonary disease and could be extreme enough to affect the nodule’s growth estimation. The overlap of demographic characteristics and risk factors in patients evaluated for CAD or PN follow-up makes this area of research clinically relevant.
A physiological model of cardiopulmonary hemodynamics should consider that the total time required for the pressure wave to reach the PN (transit time) depends on the complex interaction between vascular and hydrostatic pressures, vascular resistance, the distance the blood travels between leaving the right ventricle and reaching the nodule (vascular distance), and the cross-sectional area of the vessels along the way. All these factors may affect the transit time for every PN differently so that the pressure peak reaches each PN at a different cardiac cycle phase.
The dynamic change of the intravascular pressure felt around or inside the PN could change the blood volume included in the segmentation result and consequently the volume measurement. If the change to the segmentation volume is not consistent between follow-up scans, the VDT will be either under- or overestimated.
During PN follow-up, events such as the onset of pleural effusion related to congestive heart failure could alter the PN position, vascular distance (as the aerated lung parenchyma is displaced by the volume of pleural effusion), and intravascular pressure, potentially impacting subsequent volume measurements.
Chronic thromboembolism and vasculitis are associated with increased resistance and tortuosity of the distal pulmonary vessels, affecting the propagation of the intra-vascular pressure wave, and the pressure magnitude. Clinically, these hemodynamic changes can lead to pulmonary hypertension and ultimately right ventricular dysfunction. This phenomenon increases the diameter of the MPA in both systole and diastole, but predominantly in the later, as can be seen on CT, reducing the diameter difference between both cardiac phases. In contrast, pulmonary valvular regurgitation would increase this difference by predominantly reducing the diastolic pressure [19].
Changes in hydrostatic pressure (anteroposterior gradient in a patient in decubitus position on the CT scanner), changes in the vessels' cross-sectional luminal area (craniocaudal gradient), and the vascular distance further explain the significance of the location of the nodule in our model.
Cardiomegaly was also significant in our model but had negligible effect size. One possible explanation is the increased pressure in the left atria with retrograde capillary recruitment around the PN [20]. The potential influence of sex is also difficult to explain and has a negligible effect size but may be related to a higher vessel density in women than in men, as suggested in a recent study [21].
The volume measurement is unsurprisingly strongly related to the average diameter, as both relate to a different aspect (i.e., volume and diameter) of the same characteristic (i.e., size). Likewise, the volume is also related to the appropriateness of the nodule segmentation with a moderate effect size. In clinical practice, cases considered inadequate will tend to overestimate the volume (e.g., the inclusion of the bronchial wall in segmentation) and should be manually corrected. By including this variable in the model, we can minimize inter-observer variability by avoiding manual correction of the semiautomatic segmentation results but still be able to distinguish its effect from the effect attributable to cardiopulmonary hemodynamic factors.
As expected, the inter-observer agreement is very high globally but even higher in diastole, which could be related to minimizing cardiac motion using ECG-gating, as suggested by Boll et al. [18].
No reasonable inter-software agreement can be assumed, which exposes the differences in segmentation performance across different software packages, and reinforces the current recommendations that the same software package should be used throughout the follow-up period [22].
CCTA scans allow multiple independent measurements of a PN (at different phases of the cardiac cycle) in a single acquisition, controlling for most other factors (such as the absence of true growth). In this way, it is similar to the coffee-break study design, which also assumes a lack of true growth (zero-change datasets). However, the latter is more useful in phantom studies because of the increased radiation exposure [23]. Nevertheless, investigating the effect of these hemodynamic changes on PN volumetry would not be feasible with a coffee-break study design using non-gated scans (zero-change even for hemodynamic changes) and would be very inefficient in a longitudinal study (because we cannot assume the absence of true growth over time and the true volume of a PN is unknown).
We propose future research into the variability of the volume measurement in CCTA scans since growth estimation is the most critical application of volumetry in PN between 5 and 8 mm.
A limitation of the present study is the low percentage of cases included in the study from the large study sample, which is due to the large number of CCTA scans with missing systolic or diastolic phases in PACS. This is related to the department approach of sending only the most diagnostically relevant phases for archiving. Nevertheless, to our knowledge, this study represents the largest published series on PN volumetry in ECG-gated CT scans and using current and updated volumetry tools in clinical use. Another limitation of the study is the lack of information on risk factors such as smoking. Given the substantial overlap between lung cancer and CAD risk factors, we assume that their prevalence is high in the study population. We realize that the model simplifies the complex mechanisms of cardiopulmonary hemodynamics and does not attempt to address the systemic bronchial arteries or the regulatory mechanisms of ventilation and perfusion. Other non-hemodynamic factors may also influence the measurement, like the distribution of the nodules along the airways, which may also be confounded with the vascular distance.