Introduction

Risk stratification after acute pulmonary embolism (PE) is important because it allows for appropriate clinical management [1]. As PE elevates pulmonary vascular resistance, right ventricular afterload increases. Left unchecked, this may lead to a dilated, hypokinetic right ventricle that ultimately causes systemic hypotension and cardiac arrest [2]. Echocardiography is the established imaging modality for risk stratification after PE; several studies suggest that echocardiographic right ventricular dilatation or right ventricular wall hypokinesis may identify those at high risk for right ventricular failure and death [3–6].

Computed tomography (CT) is the first line diagnostic study for PE [7, 8]. As the entire heart is imaged on nearly every PE protocol CT, there has been substantial interest in using the diagnostic CT pulmonary angiogram (CTPA) to assess the right ventricle and by extension to assess prognosis after PE [9]. Several studies support an increased RV/LV diameter ratio, as measured in the standard axial or reformatted four-chamber views, as a predictor of short term mortality after PE [10–22], though the use of CTPA for this purpose remains controversial [23]. For patients with a negative prior examination, measuring the interval increase in the RV/LV ratio improves specificity since the patient may have preexisting RV enlargement before PE [24].

In routine imaging for patients with suspected PE, CT data is acquired without ECG gating. One reason is that pulmonary arterial pulsation artifact is usually small [25]. For visualization of the pulmonary arteries, breathing artifacts are typically more significant than pulsation artifacts [26], and ECG-gated scans generally have a longer acquisition time than comparable protocols without ECG-gating. The notable exception is single gantry rotation ECG-gated wide area detector axial CT [27]. The second and perhaps more significant reason to avoid ECG gating in CTPA is the greater patient radiation dose. Thus, from the perspective of detecting a filling defect in the pulmonary arterial system, the support for implementing ECG-gating into routine protocols is relatively weak. Moreover, CTPA is widely used because of the clinical and laboratory challenges in excluding patients with clinically suspected PE. Hence, the large majority of studies are negative for PE, and the cumulative increased radiation to those patients who do not have PE can be substantial.

When considering CTPA as an exam that not only makes the diagnosis of PE but also gives prognostic information via the RV/LV measurements, there are potential advantages of incorporating ECG-gating, despite the increased radiation. The first potential advantage is that ECG-gating minimizes and in most cases eliminates cardiac motion artifact. Thus, measurements of the ventricular diameters can be, in theory, more precise. The second advantage is that retrospective ECG-gating can be performed with CT acquisition throughout the cardiac cycle. This information enables RV cine evaluation so that wall motion and RV function can be assessed, akin to echocardiography. The purpose of this study is to prospectively compare cardiac ventricular measurements from non-gated CT and end-diastolic ECG-gated CT in patients with acute pulmonary embolism (PE).

To achieve significant results with a modest prospective patient cohort we have created a statistical model that reflects non-gated cardiac CT versus ECG-gated CTPA. Specifically, we assume that the ECG-gated exam has less motion artifact and model the motion introduced by the absence of gating as an error in measurement to test the hypothesis that ECG-gated exams have improved specificity for predicting 30-day mortality after PE.

Materials and methods

Study subjects

The institutional human research committee approved prospective ECG-gated CT dedicated to evaluate the right ventricle. Thirty adult patients (16 female, mean age = 56 years, range = 26–77 years) diagnosed with PE by routine non-gated CT were enrolled between July 2005 and February 2006 at a single institution. Each subsequently received a second, ECG-gated CT to evaluate their right ventricle within 36 h of the first non-gated CT examination.

Image acquisition

Non-gated CTPA was performed according to our standard institutional protocol. Four, sixteen and sixty-four slice multidetector (MD) CT scanners (Siemens Medical Systems, Erlangen, Germany) were used with 120 kV and an effective mAs ~200. One-hundred and twenty-five cc of iodinated contrast (Ultravist 370, Berlex, Richmond, CA) were administered via a power injector. Bolus tracking was employed on the main PA, with image acquisition at 1.0–1.25 × 0.75–1.0 mm.

The ECG-gated CT scans were performed on 16 (gantry rotation time 420 ms) and 64 (gantry rotation time 330–370 ms) slice MDCT scanners (Siemens) at 120 kV and an effective mAs of ~300–400. Fifty cc of iodinated contrast (Ultravist 370) were administered via a power injector. Bolus tracking was employed, with image acquisition at 1.5 × 1.0 mm. The 1 mm axial images were reconstructed at every 12.5% of the RR interval, yielding eight phases. The two phases corresponding to end-diastole and end-systole were visually identified and used for image post-processing and analysis.

Image post-processing and analysis

Each patient had three series of images: the non-gated diagnostic CTPA, the ECG-gated end-diastolic images, and the ECG-gated end-systolic images. For all three series, four-chamber reformations designed to standardize the RV/LV diameter ratios [28] were rendered on a dedicated 3D workstation (Leonardo Siemens, Erlangen, Germany). The left and right ventricular diameters, areas, and volumes were measured. Ventricular diameters were measured as the greatest distance between the interventricular septum and the free ventricular wall along the short axis of the heart. For the four-chamber measurements, the left and right ventricular diameters were measured on a single image best depicting the four-chamber view [20]. For the axial view, the diameter of each ventricle was measured on the axial level where it was largest [19]. Ventricular areas were measured on the same images as their respective diameters and were defined by the border of the valve plane, endocardium and interventricular septum, inclusive of papillary muscles. For the measurement of volumes, a series of consecutive 3 mm short axis images was reformatted. Endocardial contours inclusive of papillary muscle and trabeculation were manually traced then automatically summed by the workstation to calculate ventricular cavity volumes [29–32]. For the ECG-gated scans, ventricular ejection fractions were calculated from end-systolic and end-diastolic volume measurements.

Statistical analysis

For the RV/LV diameter, area and volume ratios, Spearman’s Rank Correlation Coefficients (RCC) with 0.90 confidence intervals were calculated for non-gated versus gated end-diastolic images. In addition, the median absolute differences between the gated and non-gated measurements relative to the gated measurements were provided to summarize the degree to which the two measurements differ. A statistical model estimated the potential increase in specificity of gated end-diastolic measurements compared to non-gated measurements of diameter, area and volume ratios. The model made three assumptions. The first assumption was that the log-transformed non-gated measurements were the log-transformed gated measurements plus an independent normally distributed random error, the latter attributed to motion inherent in measurements obtained from non-gated images compared to similar measurements obtained from a gated exam. The second assumption was that the RV/LV diameter ratios of patients with and without RV dysfunction follow log-normal distributions with different means but the same variance. The third assumption, extracted from the literature [20], was that that four-chamber RV/LV diameter ratios achieve a sensitivity = 0.78 and specificity = 0.38 to predict 30-day mortality (prevalence = 0.15) for a four-chamber RV/LV diameter ratio >0.9. To our knowledge, there is no CT data that associate area and volume ratios with death after PE. Thus, the area and volume ratios used the same assumptions for sensitivity and specificity. These assumed models lead to binormal ROC curves [33, 34] for both gated and non-gated measurements. We use the estimated binormal ROC curves to calculate the potential improvement in specificity attributed to using the gated end-diastolic measurements. The same analysis was performed for ventricular area and volume ratios. The bootstrap methods were used to obtain confidence intervals for all the parameters of interest [35].

Results

The rank correlation coefficients (Table 1) demonstrate a high level of correlation between non-gated versus gated end-diastolic measurements. The RCC (0.90 CI) of the axial RV/LV diameter ratio was 0.88 (0.74–0.95). The RCC of the four-chamber RV/LV diameter ratio was 0.83 (0.68–0.90). Furthermore, the median absolute percent differences (Table 2) also suggest a high degree of concordance between the gated and non-gated measurements. On average, the two measurements differ by 8.0% (0.90 CI: [5.4%, 10.6%]) for the axial RV/LV diameter ratio and by 6.8% (0.90 CI: [2.1%, 11.6%]) for the four-chamber RV/LV diameter ratio. These findings suggests limited or no potential benefit from ECG gating for the assessment of ventricular diameter ratios.

Table 1 Non-gated versus ECG-gated end-diastolic rank correlation coefficients with standard error and 0.90 confidence interval (CI)
Full size table
Table 2 Median absolute percent difference between the gated and ECG-gated end-diastolic measurements with standard error and 0.90 confidence interval (CI)
Full size table

Assuming that non-gated diameter ratios achieve a specificity of 0.38 at a sensitivity of 0.78, the predicted specificity of end-diastolic ECG-gated diameter ratios ranged from 0.42 to 0.46 (Table 3). The predicted gain in specificity ranged from 0.035 to 0.083, indicating that diameter, area, and volume measurements obtained from ECG-gated CT provides limited potential improvement over non-gated CT for predicting 30-day mortality.

Table 3 Predicted gains in specificity using ECG-gated end-diastolic ratios rather than non-gated ratios (assuming non-gated sensitivity = 0.78 and specificity = 0.38)
Full size table

The mean LV ejection fraction was 0.60 ± 0.068 (range 0.42–0.75), and the mean RV ejection fraction was 0.48 ± 0.10 (range 0.33–0.65).

Discussion

The cost of evaluating the right ventricle with a separate ECG-gated CT scan includes the administration of iodinated contrast, radiation exposure, and time. The potential benefit is ventricular measurements without motion artifact. We demonstrate only minimal incremental change from measurements of ventricular diameter, area, and volume from ECG-gated acquisition. To date, the most widely studied measurements are the axial and four-chamber diameter ratios. The high correlation between non-gated and gated end-diastolic diameter ratios suggests minimal benefit from performing ECG gating for these measurements. This is further supported by the minimal improvement in specificity for 30-day mortality using the statistical model. Our study demonstrates correlation between non-gated and gated end-diastolic area and volume measurements. The correlation was not as strong as that for the diameter ratios, probably reflecting more variability in these measurements (Figs. 1, 2).

Fig. 1
figure 1

Four-chamber views shown with the ventricular diameter measurements obtained in ECG-gated end-systole (a) (top) and end-diastole (b) (middle) in a 59-year-old man diagnosed with acute PE. (c) (bottom) four-chamber ventricular diameter measurements obtained from the non-gated CT pulmonary angiogram in the same patient

Full size image
Fig. 2
figure 2

Four-chamber views shown with the ventricular diameter measurements obtained in ECG-gated end-systole (a) (top) and end-diastole (b) (middle) in a 57-year-old woman diagnosed with acute PE. (c) (bottom) four-chamber ventricular diameter measurements obtained from the non-gated CT pulmonary angiogram in the same patient

Full size image

In a related study by Dogan et al. [36], 66 patients, 29 of whom had PE, were examined with non-gated and ECG-gated CT. The mean four-chamber RV/LV diameter ratios and RV end-systolic volumes were higher in patients with PE than those with no PE. Furthermore, end-systolic RV volumes and end-systolic and end-diastolic RV/LV volume ratios were significantly different between those patients with central versus peripheral PE. These important findings suggest that diastolic and systolic volumes help assess the severity of PE. The methods of our study differ in that we directly compared non-gated and ECG-gated RV/LV diameter ratios as predictors of mortality after PE. To date, the RV/LV diameter ratio is the only ventricular CT measurement directly associated with mortality after PE [9].

The statistical model achieved narrow confidence intervals to compare measurements with and without ECG-gating using a relatively small patient cohort. However, a limitation of this study is that this sample is too small to study the potential utility of ventricular ejection fraction and wall motion abnormalities. To explore the full potential of ECG-gated exams, morphologic parameters could be used in conjunction with right ventricular function in a large enough sample where there are a significant number of patients with 30-day mortality. In addition, dual source CT scanners achieve an 83 millisecond temporal resolution so that the cine assessment of the ventricular walls to assess the RV can more closely approximate echocardiography. The potential drawback of this method is that retrospective ECG-gating is required, increasing the radiation dose in comparison to prospective ECG gated methods.

In conclusion, ventricular measurements obtained from non-gated CT images correlate with their ECG-gated end-diastolic counterparts. The correlation is strongest with RV/LV diameter ratios. For ventricular diameters, areas, and volumes, modeling the cardiac motion inherent in non-gated CTPA as an error with respect to measurements made from ECG-gated studies shows minimal improvement in specificity for predicting 30-day mortality after PE. Larger patient samples focused on PE related mortality can more fully address the impact of incorporating ECG-gated CT into the evaluation of RV dysfunction and the tradeoffs associated with the associated radiation exposure.