Vascular and Parenchymal Enhancement Assessment by Dual-Phase Dual-Energy CT in the Diagnostic Investigation of Pulmonary Hypertension

Abstract

Purpose

To evaluate pulmonary hypertension (PH) determination by dual-phase dual-energy CT pulmonary angiography vascular enhancement and perfused blood volume (PBV) quantification.

Materials and Methods

In this prospective study, consecutive participants who underwent both right heart catheterization and dual-phase dual-energy CT pulmonary angiography were included between 2012 and 2014. CT evaluation comprised a standard pulmonary arterial phase dual-energy CT pulmonary angiography acquisition (termed series 1) followed 7 seconds after series 1 completion by a second dual-energy CT pulmonary angiography acquisition limited to the central 10 cm of the pulmonary vasculature (termed series 2). In both series, enhancement in the main pulmonary artery (PAenh), the descending aorta (DAenh), and whole-lung PBV (WLenh) was calculated from dual-energy CT pulmonary angiography iodine images. Dual-energy CT pulmonary angiography and standard cardiovascular metrics were correlated to mean pulmonary artery pressure (mPAP) and pulmonary vascular resistance (PVR) with additional receiver operating characteristic curve analysis.

Results

A total of 102 participants (median age, 70; range, 58–78 years; 60 women) were included. Sixty-five participants had PH defined by mPAP of greater than or equal to 25 mm Hg, and 51 participants had PH defined by PVR of greater than 3 Wood units. By either definition, participants with PH had higher PAenh/WLenh ratio and lower WLenh and DAenh in series 1 (P < .05) and higher PAenh and WLenh in series 2 (P < .05). Change in WLenh determined highest diagnostic accuracy to define disease by mPAP (area under the receiver operating characteristic curve [AUC], 0.78) and PVR (AUC, 0.79) and the best mPAP correlation (r = 0.62). PAenh series 2 correlated best with PVR (r = 0.49). Multiple linear regression analysis incorporating WLenh and series 1 DAenh improved PVR correlation (r = 0.56). Combining these dual-energy CT pulmonary angiography metrics with main pulmonary artery size and right-to-left ventricular ratio achieved the highest correlations (mPAP, r = 0.71; PVR, r = 0.64).

Conclusion

Dual-phase dual-energy CT pulmonary angiography enhancement quantification appears to improve mPAP and PVR prediction in noninvasive PH evaluation.

Supplemental material is available for this article.

See also the commentary by Kay in this issue.

© RSNA, 2020

Summary

Dual-phase dual-energy CT may further the noninvasive assessment of pulmonary hypertension by examining the change in volumetric whole-lung enhancement after 7 seconds and/or delayed pulmonary artery enhancement.

Key Points

• ■ Dual-phase dual-energy CT pulmonary angiography vascular enhancement and perfused pulmonary blood volume quantification is significantly different in participants with and without pulmonary hypertension.

• ■ Change in volumetric whole-lung enhancement after 7 seconds delivers good diagnostic accuracy to identify pulmonary hypertension (area under receiver operating characteristic curve, 0.78–0.79).

• ■ The highest correlation with mean pulmonary arterial pressure was achieved by augmenting standard cardiovascular metrics in clinical use (right-to-left ventricular ratio and main pulmonary artery–to–ascending aorta ratio) with the change in whole-lung enhancement metric from dual-phase dual-energy CT pulmonary angiography (r = 0.71).

Introduction

Pulmonary hypertension (PH) is a life-threatening disease in which early diagnosis improves poor prognosis (1). A diagnosis of PH requires invasive right heart catheterization (RHC) to confirm a mean pulmonary arterial pressure (mPAP) of greater than or equal to 25 mm Hg at rest (2). More recently, the clinically important parameter pulmonary vascular resistance (PVR, > 3 Wood units [WU]), which is obtained by RHC, is now required to diagnose the group 1 clinical subset (pulmonary arterial hypertension) (2).

PH is a heterogeneous condition that can be idiopathic or associated with many other conditions, such as cardiorespiratory disease and connective tissue disease (1). In the investigation of PH, CT is traditionally used to help identify associated lung disease or with the use of CT pulmonary angiography for the presence of chronic thromboembolic PH (CTEPH). Historically, the identification of PH on the basis of CT imaging is limited, whereby a diagnosis can be suggested from cardiovascular metrics, including main pulmonary artery (MPA) size or the disproportionate enlargement of the right ventricle (RV) (3).

Smaller series retrospective work has suggested that assessment of pulmonary perfused blood volume (PBV), using iodine images acquired from dual-energy CT pulmonary angiography, could improve the noninvasive diagnosis of PH (4,5). Main pulmonary artery enhancement (PAenh), when compared with the volumetric enhancement of the whole lung (WLenh) expressed as a ratio PAenh/WLenh, significantly correlated with the invasive hemodynamic parameter PVR (4). This ratio is a proxy for high central-to-peripheral pulmonary PBV in patients with increased resistance to pulmonary blood flow.

It has since been suggested that PBV defects detected at dual-energy CT pulmonary angiography correlate with mPAP and PVR and can be used to guide and monitor therapeutics in patients with CTEPH (6,7). There has also been initial interest in dual-phase dual-energy CT pulmonary angiography studies, currently limited to CTEPH (8,9).

The aim of this prospective, RHC-defined study was to assess vascular and parenchymal enhancement assessment by PBV to diagnose PH in a heterogeneous cohort using dual-phase dual-energy CT pulmonary angiography. Standard CT metrics to identify PH were also studied for this participant population. We hypothesized that the addition of a second phase could advance the use of pulmonary PBV as a surrogate for lung perfusion by offering a further time point for assessment. We additionally hypothesized that the presence of bronchial collaterals may be a variable that may account for delayed lung perfusion.

Materials and Methods

Study Design

Consecutive participants referred for RHC to assess PH were prospectively recruited with informed consent to this institutional review board–approved single-center cohort study (study approval reference 12/WA/0253) between 2012 and 2014. A total of 106 participants met eligibility criteria. Participants were excluded if they did not end up undergoing RHC. A total of 102 participants underwent RHC and dual-phase dual-energy CT pulmonary angiography, defined by sample size calculation, setting α to .05, power to 0.8, and expecting 50% disease rate based on local audit data (10). RHC was performed via the internal jugular vein using standard procedure with thermodilution technique (11,12).

Dual-Energy CT Pulmonary Angiography Protocol

A standard dual-energy CT pulmonary angiography protocol was performed with tube voltages of 100 and Sn140 kVP and a quality reference mAs of 150/120 with caudocranial acquisition with a Definition Flash scanner (Siemens Medical Solutions, Forchheim, Germany). Participants were given 100 mL of iohexol (300 mg iodine per milliliter) at 5 mL per second (GE Healthcare, Waukesha, Wis) with bolus tracking at a 100 HU threshold. Data acquired from these protocols were termed series 1. This series was supplemented by a limited (10-cm z-axis coverage scan, centralized to MPA) central systemic arterial phase dual-energy CT pulmonary angiography of the pulmonary arteries, commencing 7 seconds after series 1 completion (with tube voltages of 100 and Sn140 kVP, quality reference mAs of 150/128, and caudocranial acquisition) which was termed series 2. The 7-second interval was empirically selected to be as soon as practicable following the first acquisition, permitting table repositioning and participant respiration before the second breath-hold acquisition. Allowing for time of scan acquisitions, this resulted in an approximate 11–12-second interval between the center timepoint of the pulmonary arterial and systemic arterial acquisitions. Analysis using only a partial series was proven comparable in preliminary work to a full coverage series and used to reduce radiation exposure and maximize applicability to clinical practice (Appendix E1 [supplement]).

Data Processing and Analysis

All data were postprocessed on a workstation by a single reader (Multimodality Workplace, Siemens Medical Systems). PBV maps were automatically generated from the dual-energy CT pulmonary angiography images using the “Lung PBV” algorithm on a commercially available postprocessing software platform (Syngo Via, Siemens Healthcare), setting minimum and maximum lung attenuation thresholds as −960 to −300 HU.

For each examination, complete volumetric data sets (1-mm thickness at 0.8-mm interval) were reconstructed using a soft convolution algorithm (D30) at 100 kVp, Sn140 kVp, and a weighted average set (with 40%:60% weighting from the Sn140:100 kVp images respectively).

Two-dimensional quantitative analyses were performed using the volumetric PBV axial image data, reviewed on the workstation. A 15–20-mm diameter circular region of interest measurement of iodine enhancement was recorded over the central pulmonary artery (PAenh, [Fig 1a]) and descending aorta (DAenh) (at the level of pulmonary artery bifurcation) for both series 1 and 2. Three-dimensional quantitative analyses were performed using automated three-dimensional volumetric lung segmentation (Syngo Via) for series 1 and 2 with measurements of the mean volumetrically determined iodine enhancement of both lungs (WLenh, Fig 1b). PAenh divided by WLenh was calculated as the ratio of central-to-peripheral enhancement (PAenh/WLenh) for both series. These methods were in accordance with previous published work (4).

Figure 1a:

(a) Dual-energy CT image (series 2) displaying the region of interest chosen over the pulmonary artery at the level of the pulmonary artery bifurcation with automatic display of the level of enhancement. (b) Dual-energy CT pulmonary blood volume map for a participant without pulmonary hypertension. The automated numeric display of whole-lung enhancement (circled) is shown.

Figure 1b:

(a) Dual-energy CT image (series 2) displaying the region of interest chosen over the pulmonary artery at the level of the pulmonary artery bifurcation with automatic display of the level of enhancement. (b) Dual-energy CT pulmonary blood volume map for a participant without pulmonary hypertension. The automated numeric display of whole-lung enhancement (circled) is shown.

Two reviewers (C.S. and S.S. with 5 years and 4 years of experience in thoracic imaging, respectively) performed standard cardiovascular CT measurements using electronic calipers while blind to clinical details. Review of series 1 contiguous 1-mm thickness axial data sets on mediastinal windows on a picture archiving and communication system was performed. The MPA short-axis size was recorded at its widest point, perpendicular to the long axis of the vessel, at least 1 cm proximal to the trunk bifurcation. The ascending aorta (AA) measurement was performed at the same level as MPA measurement and the ratio between them calculated (MPA/AA aorta) (13–15). Maximum RV and left ventricle (LV) diameters were defined manually as the maximum distance from the interventricular septum to endocardial border, typically at different craniocaudal levels, and their ratio was calculated (RV/LV) (16). If quantitative CT measurements were discrepant by more than 20% between readers, a third reader (I.V., with 20 years of experience in thoracic imaging) provided a further independent measurement, and then a consensus measurement was reached between the three readers.

To assess for the potential influence that the presence of bronchial collaterals could have on the studied dual-energy CT pulmonary angiography variables, further analysis was performed. Two thoracic radiologists (I.V. and A.D., with 20 years and 15 years of experience, respectively), blinded to participant diagnosis and metrics, classified by consensus the presence of bronchial collaterals on a three-point scale according to combined size, extent, and tortuosity (0 = normal, 1 = mild, 2 = moderate, and 3 = markedly prominent).

Statistical Analysis

Statistical analyses were performed using Statistical Package for the Social Sciences (SPSS 21.0 for Windows; SPSS, Chicago, Ill). Statistical significance was defined as P < .05. Calculated dual-energy CT pulmonary angiography metrics were analyzed individually within one series, and the change from series 1 to 2 was calculated and analyzed. For all analyses, participants were grouped according to the defined thresholds for a diagnosis of PH of mPAP of greater than or equal to 25 mm Hg, then for a PVR of greater than 3 WU (2).

Potentially significant differences in CT variables between groups defined by mPAP (< or ≥ 25 mm Hg), then PVR (≤ or > 3 WU) (2), were analyzed by an independent samples t test. If a positive Levene test determined unequal variance, then the separate variances t test was used instead.

Diagnostic accuracy for the parameters to determine mPAP greater than or equal to 25 mm Hg, then PVR greater than 3 WU, were assessed using nonparametric receiver operating curve analysis to calculate area under the receiver operating characteristic curve (AUC).

The direction and strength of univariate associations between CT pulmonary angiography-derived parameters and RHC-derived mPAP and PVR were determined by scatter plot evaluation and Pearson correlation coefficients (r) and confirmed with entry into a stepwise regression. Where expressed, r values reflect unadjusted r values. The natural log of the variable was used when the one-sample Kolmogorov-Smirnov test was statistically significant, indicating the data were otherwise not normally distributed.

To determine whether combining dual-energy CT pulmonary angiography metrics strengthened their ability to predict mPAP or PVR, a multiple linear regression analysis was performed. Based on the strength of their univariate correlation, and the stepwise regression, dual-energy CT pulmonary angiography variables were entered into a multiple linear regression analysis to predict mPAP and then PVR. The best model was chosen on statistical criteria.

To analyze the dual-energy CT pulmonary angiography results alongside the current standard conventional CT pulmonary angiography metrics to identify the presence of PH, further multiple linear regression analyses were undertaken. The best standard CT pulmonary angiography metrics were combined with the best dual-energy CT pulmonary angiography variables to predict mPAP and then PVR in a multiple linear regression analysis and the statistically best model chosen.

The incidence of significant bronchial collaterals was compared in participants with increasing delayed lung enhancement (change in WLenh increased) to that of participants with reduced delayed lung enhancement (change in WLenh decreased) to determine if this additional flow was bronchial derived. The three-point scale bronchial collateral score was entered with the best dual-energy CT pulmonary angiography variables into the multiple linear regression analysis to see if it entered the equation to predict PVR.

Results

Dual-energy CT pulmonary angiography and RHC were performed in 102 participants (median age, 70 years; range, 58–78 years; 60 women) across all PH clinical classification groups (2) (Table 1). Four additional recruited participants underwent dual-energy CT pulmonary angiography but did not later undergo RHC (three for clinical reasons and one due to participant choice) and were excluded from the study (Fig 2). A total of 64% (65 of 102) of participants had PH defined by mPAP of greater than or equal to 25 mm Hg and 50% (51 of 102) of participants had elevated PVR of greater than 3 WU. The range of time between RHC and dual-energy CT pulmonary angiography was 0–54 days, median 6 days.

Table 1:

Clinical Characteristics of Participants

Figure 2:

Study flow diagram showing the number of participants recruited and undergoing right heart catheterization (RHC), dual-phase dual-energy CT pulmonary angiography (DE-CTPA), and the number of participants in each hemodynamic group defined by mean pulmonary artery pressure (mPAP) and then pulmonary vascular resistance (PVR).

All dual-energy CT pulmonary angiography and standard CT pulmonary angiography cardiovascular metrics were calculated successfully in all participants except one with PH who had no contrast material detectable in the descending aorta in series 1, therefore no measurement for DAenh was undertaken. The median dose-length product and effective radiation dose were 396 mGy · cm (range, 310–490 mGy · cm) and 5.5 mSv (range, 4.3–6.9 mSv), respectively, for the full CT examination with dual series. The median dose-length product and effective radiation dose for the additional image series alone were 94 mGy · cm and 1.3 mSv, respectively.

Comparison between Groups

Significant differences were identified in the dual energy–determined enhancement characteristics between participants with elevated mPAP or PVR and those with normal values. The relationships in the contrast enhancement for both image series are shown in Table 2

Table 2:

Comparison of Mean Dual-Energy CT Pulmonary Angiography Parameters between Series 1, Series 2, and Standard CT Metrics

Series 1: Standard pulmonary arterial phase.— Participants with PH had lower WLenh compared with participants without PH, with a higher PAenh/WLenh ratio (when defined by mPAP, P = .002; and PVR, P < .001). The DAenh were lower in participants with PH (defined by mPAP, P < .001; and PVR, P = .01) when compared with patients without PH.

Series 2: Delayed phase.— By the second image series, PAenh was higher in the participants with PH compared with participants without PH (defined by mPAP, P < .001; and PVR, P < .001). The WLenh measured in the delayed series was also higher in participants with PH when compared with participants without PH (when defined by mPAP, P = .04; and PVR, P < .001). The PAenh/WLenh ratio in participants with PH remained elevated at the delayed time point (defined by mPAP, P = .02; and PVR, P = .005).

Change between series.— PAenh predictably decreased in all participants from series 1 to delayed series 2. For participants with PH defined by mPAP, this decrease was less than that for participants without PH (P = .02). A similar PAenh trend was observed for participants with PH defined by PVR, but this did not attain statistical significance.

The mean WLenh decreased from series 1 to 2 in participants without PH whether defined by mPAP (−9.4 HU) or PVR (−8.0 HU). Conversely, for both mPAP- and PVR-defined PH groups, the mean change in WLenh was positive from series 1 to 2 (mPAP, +1.5 HU; and PVR, +3.1 HU, Fig 3). This magnitude in change of WLenh for participants with PH was different compared with participants without PH (P < .001).

Figure 3:

Boxplot to compare the change in whole-lung enhancement in groups defined by pulmonary vascular resistance (PVR) (pulmonary vascular pressure ≤ or > 3 Wood units [WU]).

Standard conventional CT metrics.— MPA diameter, MPA/AA ratio, and RV/LV ratios were all different between groups defined by mPAP (P = .01 to P < .001) or PVR (P < .001).

Diagnostic Accuracy

The change in WLenh from series 1 to 2 was the metric with the highest diagnostic accuracy delivered by dual-energy CT pulmonary angiography for identifying participants with PH shown by AUC analysis (Table 3 and Fig 4). The AUC for this dual-energy CT pulmonary angiography metric to predict PH defined by mPAP of greater than or equal to 25 mm Hg was 0.78 (95% CI: 0.69, 0.87) and to predict PVR of greater than 3 WU was 0.79 (95% CI: 0.71, 0.88) (Table 3 and Fig 4).

Table 3:

AUC for Parameters Predicting Pulmonary Hypertension

Figure 4a:

(a, b) The receiver operating characteristic curves (ROCs) are displayed for the highest area under the curve for predicting disease for both standard CT measurements and dual-energy parameters, defined by mPAP (a) then PVR (b). Additional curves are provided for best composite measures defined by multivariate linear regression analysis incorporating both dual-energy metrics and conventional CT morphologic metrics. mPAP = mean pulmonary artery pressure, PAenh2 = pulmonary artery enhancement series 2, PVR = pulmonary vascular resistance, WLenh = whole lung enhancement.

Figure 4b:

(a, b) The receiver operating characteristic curves (ROCs) are displayed for the highest area under the curve for predicting disease for both standard CT measurements and dual-energy parameters, defined by mPAP (a) then PVR (b). Additional curves are provided for best composite measures defined by multivariate linear regression analysis incorporating both dual-energy metrics and conventional CT morphologic metrics. mPAP = mean pulmonary artery pressure, PAenh2 = pulmonary artery enhancement series 2, PVR = pulmonary vascular resistance, WLenh = whole lung enhancement.

Overall, the best CT pulmonary angiography parameter to detect PH defined by mPAP was MPA size (AUC 0.80 [95% CI: 0.71, 0.89]), whereas the dual-energy CT pulmonary angiography parameter change in WLenh showed the highest diagnostic accuracy to predict PVR of greater than 3 WU, above all CT pulmonary angiography metrics studied (Table 3 and Fig 4).

Unadjusted Single Parameter Correlation

The best CT pulmonary angiography parameter that correlated with mPAP was the change in WLenh from series 1 to 2 (r = 0.62, P < .001) (Table 4 and Fig 5). This correlation was higher than the standard CT metrics (MPA diameter, r = 0.54; MPA/AA ratio, r = 0.45; and RV/LV ratio, r = 0.46 [P < .001]).

Table 4:

Univariate Correlation with Dual-Energy CT Pulmonary Angiography

Figure 5:

Scatterplot shows correlation between mean pulmonary artery pressure and change in whole-lung enhancement from standard CT pulmonary angiography series (series 1) to delayed image series (series 2) (r = 0.62).

PVR was log transformed for further analysis (Ln PVR) as it did not meet the statistical criteria for normal distribution. Overall, the CT pulmonary angiography parameter with the highest correlation with Ln PVR was RV/LV ratio (r = 0.53, P < .001) (Table 4). The dual-energy CT pulmonary angiography parameters that correlated best with Ln PVR were PAenh series 2 (r = 0.49, P < .001) and change in WLenh from series 1 to 2 (r = 0.47, P < .001). These correlations were higher than found for MPA diameter (r = 0.34). The correlations between Ln PVR and series 1 dual-energy CT pulmonary angiography metrics were also lower, with series 1 PAenh/WLenh demonstrating only modest correlation with Ln PVR (r = 0.34, P < .001).

Multiple Linear Regression Analysis

Multiple linear regression analysis identified the best dual-energy CT pulmonary angiography metrics in single or combined image series to predict mPAP (Table 5), then PVR (Table 6). To predict average mPAP, the single metric change in WLenh delivered the best model (r = 0.62), with the following formula:

Table 5:

Multiple Linear Regression Analysis to Predict Average mPAP

Table 6:

Multiple Linear Regression Analysis to Predict Average PVR

mPAP = 32.4 + (0.723 × change in WLenh), with 95% CIs for WLenh of 0.55, 0.92.

When this method was repeated to determine the best dual-energy CT pulmonary angiography metrics in single or combined image series to predict average Ln PVR, the following formula was found (r = 0.56):

Ln PVR = 0.134 + (0.043 × change in WLenh) + (0.004 × PAenh series 2) + (0.004 × DAenh series 1), with 95% CIs for WLenh of 0.02, 0.07; PAenh of 0.001, 0.008; and DAenh of 0.001, 0.007.

When the standard CT metrics alone were entered into multiple linear regression analysis to predict mPAP, this identified an increased r value by combining MPA diameter and RV/LV ratio (r = 0.61). When the standard CT metrics to predict PH were entered with the best dual-energy CT pulmonary angiography metrics, the following formula best predicted average mPAP (r = 0.71):

mPAP = 3.39 + (0.58 × change in WLenh) + (18.6 × MPA/AA ratio) + (10.6 × RV/LV ratio), with 95% CIs for WLenh of 0.40, 0.76; MPA/AA ratio of 6.2, 29.8; and RV/LV ratio of 3.0, 19.5.

Multiple linear regression analysis using all standard CT pulmonary angiography metrics did not improve upon univariate analyses for Ln PVR (RV/LV ratio, r = 0.53). However, when the standard CT metrics to predict PH were entered with the best dual-energy CT pulmonary angiography metrics, the following formula best predicted average Ln PVR (r = 0.64):

Ln PVR = −1.34 + (0.006 × PAenh series 2) + (1.59 × RV/LV ratio), with 95% CIs for PAenh of 0.003, 0.008; and RV/LV ratio of 1.01, 2.14.

Bronchial Collateralization

To assess for the potential influence that the presence of bronchial collaterals could have on the studied dual-energy CT pulmonary angiography variables to predict PVR, a further multiple linear regression analysis was performed, entering bronchial collateral size, extent, and tortuosity scored on a three-point scale (1, n = 20; 2, n = 10, and 3, n = 4). The presence of bronchial collaterals of any extent (≥ 1) was more common in participants with PH defined by PVR rather than in participants without PVR-defined PH (P = .04, χ² test). Indeed, when entered with the dual-energy CT pulmonary angiography variables into the multiple linear regression analysis, the bronchial collateral score was found to be a significant variable, entering the best prediction equation for Ln PVR (β = .26, P = .003). However, the presence of moderate or markedly prominent collaterals (score ≥ 2) was not more common in participants with increasing WLenh in series 2 compared with those with decreasing WLenh in series 2 (P = .71, χ² test).

Discussion

In this study, we examined the effectiveness of dual-phase dual-energy CT pulmonary angiography vascular and parenchymal enhancement assessment by PBV analysis to determine the presence and hemodynamic severity of PH in a large heterogeneous population under investigation for the disease. In the conventional standard dual-energy CT pulmonary angiography phase (series 1), participants with PH defined by mPAP or PVR had equivalent central PAenh but significantly lower pulmonary parenchymal and systemic aortic enhancement (WLenh, DAenh, respectively) than participants without PH. This reflects an increased central and reduced peripheral enhancement in participants with high mPAP or PVR. Expressed as the ratio PAenh/WLenh, this was the single best standard-phase dual-energy CT pulmonary angiography metric to correlate with PVR. This is in concordance to the smaller retrospective study by Ameli-Renani et al (4), although the correlation was less strong (r = 0.34 vs 0.59). These data support the hypothesis that the pulmonary transit time from central to peripheral pulmonary arteries is slower in participants with PH, which is corroborated by studies using MRI (17–20) and in animal studies (21).

In our heterogeneous population, a significant negative correlation was identified between standard-phase dual-energy CT pulmonary angiography–derived WLenh and mPAP. This is in keeping, albeit a less strong relationship, with a comparable analysis in a prior smaller study of 25 patients with CTEPH (r =−0.39 vs −0.57) (22). In both studies, WLenh alone does not appear to demonstrate a significant correlation with PVR.

These variations according to study population may be important. It is notable that in this diverse population, mPAP correlation with dual-energy CT pulmonary angiography–derived parameters (DAenh, r = −0.46; WLenh, r = −0.39; PAenh/WLenh, r = 0.38) appeared inferior to conventional metrics (main pulmonary artery, r = 0.54; RV/LV ratio, r = 0.46) on standard-phase CT pulmonary angiography. On this same phase, PAenh/WLenh correlation to PVR (r = 0.34) appeared equivalent to MPA size (r = 0.34) but still inferior to RV/LV ratio (r = 0.53) alone.

The additional delayed phase 7 seconds following the initial peak pulmonary arterial phase further highlighted the pulmonary vascular transit delay in participants with PH. In series 2, participants with PH exhibited significantly higher PAenh and WLenh than participants without PH. The average contrast enhancement in the whole lung decreased in non-PH participants by the delayed series but increased in participants with PH.

Accordingly, the highest accuracy to determine a diagnosis of PH defined by mPAP of greater than or equal to 25 mm Hg was by examining the change in WLenh between the two image series (AUC = 0.78). This appeared an equivalent diagnostic tool to diagnose elevated mPAP compared with conventional MPA size (AUC = 0.80) but superior to standard RV/LV ratio (AUC = 0.64) used commonly in clinical practice to suggest PH. Of note, the AUC for MPA size to predict elevated mPAP was lower than found in previous studies (AUC 0.87 in meta-analysis [23]), likely reflecting our heterogeneous study population.

Equally important to determining disease state defined by threshold values is to be able to characterize disease severity by a relationship of parameters to physiologic metrics. The dual-series dual-energy CT pulmonary angiography parameter change in WLenh predicted disease severity defined by mPAP better than all other metrics, including conventional standard CT parameters (correlation, r = 0.62). No combination of dual-energy CT pulmonary angiography parameters improved this single parameter correlation with mPAP. However, the addition of conventional metrics strengthened the correlation further in multiple linear regression analysis (r = 0.71). This suggests that using the dual series and determining the change in WLenh could not only help correctly identify patients with PH, but could also suggest disease severity, complemented by conventional CT metrics. Future studies may evaluate whether such metrics have the potential to risk stratify patients or guide therapeutics.

Although historically mPAP determination has been a strong focus of CT evaluation in PH, accurately determining both mPAP and PVR are critical parameters in the identification and treatment of patients with PH (3). Significantly elevated PVR is an important parameter to define true pulmonary vascular disease, now included in the diagnosis of a clinically important class of patients with PH who respond more favorably to pulmonary vasodilator therapies, many of which are expensive and may be nonefficacious or detrimental in other PH groups. The dual-series dual-energy CT pulmonary angiography parameter change in WLenh had the highest diagnostic accuracy of all dual-energy CT pulmonary angiography and conventional metrics for determining significantly raised PVR greater than 3 WU (AUC = 0.79).

PVR is accurately determined by RHC and not well determined by current noninvasive techniques such as echocardiography and CT. Therefore, the prospect of a noninvasive CT process to monitor both mPAP and PVR has substantial potential clinical value. Like previous CT studies, all parameters did not achieve the strength of correlation to PVR that is seen with mPAP. The highest correlation with PVR was examining the RV/LV ratio (r = 0.53), followed by PAenh series 2 (r = 0.49) and change in WLenh (r = 0.47). A combination of three dual-energy CT pulmonary angiography metrics improved correlation further (r = 0.56), but comparable to mPAP the best correlation to PVR was achieved by a combination of dual-energy CT pulmonary angiography and conventional metrics (PAenh2 and RV/LV ratio, r = 0.64). This highlights the potential additive value of this dual series metric to standard CT parameters used in clinical practice.

The increased delayed phase pulmonary enhancement identified in participants with PH by WLenh may be partly derived from the systemic collateral circulation by bronchial collaterals (24). Previous dual-energy CT pulmonary angiography studies have hypothesized that the presence and significance of bronchial collaterals may be an important determinant of increased delayed flow in CTEPH compared with acute pulmonary emboli (8).

Our results support that bronchial collaterals are more common in PH and may be important in PVR determination (25). However, we found that the incidence of larger bronchial collaterals was not higher in participants with increased delayed WLenh compared with those with no increase in this phase. The delayed aortic enhancement supplying the bronchials was comparable in participants with and without PH, perhaps due to rapid clearance in participants without PH and delayed enhancement in participants with PH. Despite this, the PAenh in the delayed series was significantly higher in participants with PH. Therefore, we believe that principally pulmonary vascular resistance delays pulmonary transit and results in delayed increased lung enhancement in PH, rather than bronchial blood supply increasing enhancement later. Previous studies (8) have used a longer time delay for the second series (20 seconds) which may have potentially increased the contribution of bronchial collateral flow in those cohorts.

The use of dual-energy CT pulmonary angiography techniques for PH presupposes the availability of dual-energy CT capable scanners. However, the prevalence of such technology has increased as vendors provide differing multi-energy CT solutions. Although some initial expertise is required, automated dual-energy CT pulmonary angiography PBV maps may be generated in as little as 30 seconds and provide a reader-independent tool for the assessment of global and regional lung perfusion. Dual-energy CT pulmonary angiography lacks the known risks associated with invasive RHC (3); however, the use of a dual-phase protocol does increase radiation dose exposure. In this study, this dose was minimized using reduced z-axis coverage for the second phase. Further dose reduction could be achieved using iterative reconstruction across both series, which was not available at the time of the study.

It should also be reinforced that dual-energy CT, even with the addition of a second time point limited coverage phase, is not a true analysis of perfusion, but rather a surrogate study evaluating enhancement. Perfusion CT requires multiple time point CT acquisitions evaluated by dedicated software to generate comprehensive metrics of perfusion. Indeed, our results may have varied if a longer time interval was selected between phases, or if attempts were made to adapt the timing of the second phase to individual hemodynamic variations. However, the potential value of the current technique is that it reflects only a minor modification of current dual-energy CT technique, is standardized to aid implementation, provides comprehensive chest coverage for simultaneous diagnosis, and is acquired at reduced dose compared with CT perfusion imaging.

The study was limited by the study population who were suspected of having PH and referred to a tertiary PH center site, justifying the inherent risks of RHC. The results are, therefore, applicable only to patients with a high clinical suspicion of disease. An advantage of the study was the inclusion of a broad selection of possible PH etiologies. However, an associated limitation was that the study was not powered to evaluate whether PBV measures were technically better achieved or better correlated to hemodynamic metrics within different subsets of patients with PH. Preliminary analyses suggested that this might be the case in patients with Group 1 disease, but this would require corroboration in future prospective studies. The broad inclusion was designed to best reflect a clinically relevant heterogeneous population to guide clinical practice. However, the inclusive nature of the study may have reduced the diagnostic accuracy delivered.

In conclusion, this current large prospective RHC-corroborated study determined that single-phase dual-energy CT pulmonary angiography parameters did not justify the use of dual-energy CT pulmonary angiography over standard CT, as they did not deliver higher diagnostic accuracy or correlation to pulmonary hemodynamics. However, with the addition of a second image series, improved diagnostic accuracy and significant correlation with disease hemodynamic severity was noninvasively achieved by dual-energy CT pulmonary angiography parameters. This was predominantly by the incorporation of change in whole-lung enhancement over time to diagnose PH and by the use of this parameter and delayed pulmonary arterial enhancement to characterize severity according to mPAP and PVR. Of importance, dual-energy dual-phase metrics and conventional metrics appear complementary, improving each other’s ability to predict mPAP and PVR. This suggests a potential future role for the discussed techniques, particularly in the monitoring of disease, where patients can be used as their own control, perhaps guiding PH therapeutics.