Optimal Attenuation Threshold for Quantifying CT Pulmonary Vascular Volume Ratio

Abstract

Objective

To evaluate the effects of attenuation threshold on CT pulmonary vascular volume ratios in children and young adults with congenital heart disease, and to suggest an optimal attenuation threshold.

Materials and Methods

CT percentages of right pulmonary vascular volume were compared and correlated with percentages calculated from nuclear medicine right lung perfusion in 52 patients with congenital heart disease. The selected patients had undergone electrocardiography-synchronized cardiothoracic CT and lung perfusion scintigraphy within a 1-year interval, but not interim surgical or transcatheter intervention. The percentages of CT right pulmonary vascular volumes were calculated with fixed (80–600 Hounsfield units [HU]) and adaptive thresholds (average pulmonary artery enhancement [PA_avg] divided by 2.50, 2.00, 1.75, 1.63, 1.50, and 1.25). The optimal threshold exhibited the smallest mean difference, the lowest p-value in statistically significant paired comparisons, and the highest Pearson correlation coefficient.

Results

The PA_avg value was 529.5 ± 164.8 HU (range, 250.1–956.6 HU). Results showed that fixed thresholds in the range of 320–400 HU, and adaptive thresholds of PA_avg/1.75–1.50 were optimal for quantifying CT pulmonary vascular volume ratios. The optimal thresholds demonstrated a small mean difference of ≤ 5%, no significant difference (> 0.2 for fixed thresholds, and > 0.5 for adaptive thresholds), and a high correlation coefficient (0.93 for fixed thresholds, and 0.91 for adaptive thresholds).

Conclusion

The optimal fixed and adaptive thresholds for quantifying CT pulmonary vascular volume ratios appeared equally useful. However, when considering a wide range of PA_avg, application of optimal adaptive thresholds may be more suitable than fixed thresholds in actual clinical practice.

INTRODUCTION

In patients with cyanotic congenital heart disease and peripheral pulmonary stenosis, cardiothoracic computed tomography (CT) is commonly used for the evaluation of pulmonary artery morphology, a necessity for treatment planning and post-treatment surveillance (1,2). Lung perfusion scintigraphy has been useful for assessing the differential right-to-left lung perfusion ratio, which is also essential in these patients (3,4). However, cardiothoracic CT had not been utilized to quantify pulmonary vascularity until recent studies demonstrated it could be applied to evaluate the pulmonary vascular volume ratio, in addition to pulmonary artery anatomy (5,6,7). Nonetheless, these studies demonstrated that confounding factors such as pulmonary vascular obstruction and pulmonary regurgitation increased CT measurement errors (5). CT measurements are also substantially influenced by the cardiac phase, and therefore, acquisition of cardiothoracic CT data at the end-systolic phase is recommended (6).

Additionally, the optimal attenuation threshold should be used to achieve a better result for threshold-based quantification (8,9,10,11,12,13,14). A threshold-based segmentation approach is also used to quantify CT pulmonary vascular volume ratios; therefore attenuation thresholds for quantification may influence results (5,6,7). However, the ratio may not significantly change despite different attenuation thresholds when the degree of pulmonary vascular enhancement is similar on both sides. To date, attenuation threshold effects on CT pulmonary vascular volume ratios have not been systemically assessed. Consequently, the objectives of this study were to evaluate the effects of attenuation threshold on CT pulmonary vascular volume ratios in children and young adults with congenital heart disease, and to suggest optimal attenuation thresholds.

MATERIALS AND METHODS

Our Institutional Review Board approved this retrospective study and waived the requirement for informed consent.

Study Population

The study subjects included 52 consecutive children and young adults (median age, 4 years; age range, 2 months to 28 years; 31 male patients) with congenital heart disease between January 2011 and December 2016, who underwent: 1) cardiothoracic CT examination acquired at the end-systolic phase, and 2) lung perfusion scintigraphy within a 1 year-interval (median, 60 days; range, 0–361 days), but not interim surgical or transcatheter interventions. The same study population was previously used to evaluate the clinical value of CT pulmonary vascular volume ratios (5). The majority of the patients (71.2%, 37/52) were either repaired tetralogy of Fallot (n = 27), or repaired pulmonary atresia (n = 10). The study population also had a history of surgical pulmonary artery angioplasty (n = 21), pulmonary artery stent placement (n = 4), and balloon pulmonary artery angioplasty (n = 3).

Cardiothoracic CT

A second-generation dual-source CT scanner (SOMATOM Definition Flash; Siemens Healthineers, Forchheim, Germany) was used to acquire electrocardiography (ECG)-synchronized scans from the thoracic inlet to the first lumbar vertebra in 2 × 64 × 0.6 mm slices. The z-flying focal spot technique with a 0.75-mm slice width and a 0.4-mm reconstruction interval were used. Depending on patient age and cooperation, prospectively ECG-triggered sequential scanning with additional respiratory triggering was performed in 34 free-breathing sedated patients, single breath-hold retrospectively ECG-gated spiral scanning was performed in 17 non-sedated patients, and prospectively ECG-triggered high-pitch spiral scanning was performed in one free-breathing sedated patient. Oral chloral hydrate (50 mg/kg) was initially used to sedate the patients, and intravenous midazolam (0.1 mg/kg) or ketamine (1 mg/kg) was additionally administered as required. As previously described, an additional respiratory gating system (AZ-733V; Anzai Medical Co. Ltd., Tokyo, Japan) was utilized for combined prospective ECG and respiratory triggering to minimize respiratory motions in free-breathing children (15,16). CT protocols were optimized with the following parameters to minimize radiation dose and maintain diagnostic image quality: 1) centering the patient in the CT gantry isocenter, 2) using combined tube current modulation (CARE Dose 4D; Siemens Healthineers), 3) entering the individual body size-adapted radiation dose before scan range adjustment, 4) using an aggressive version of the ECG-controlled tube current modulation (MinDose; Siemens Healthineers), 5) using a biphasic chest pain protocol for the retrospectively ECG-gated spiral scan, and 6) using sinogram-affirmed iterative reconstruction (SAFIRE; Siemens Healthineers) (17,18,19,20,21,22,23). An optimized triphasic or quadriphasic injection protocol developed at our institution was used to maximize uniform cardiovascular enhancement, and minimize peri-venous streak artifacts from undiluted contrast agent. This protocol involved an iodinated contrast agent (iomeprol; Iomeron 400, 400 mg I/mL; Bracco Imaging SpA, Milan, Italy; 1.5–2.0 mL/kg) injected at a rate of 0.5–3.0 mL/s, and a bolus-tracking technique determined scan delay, with a 150 Hounsfield units (HU) trigger threshold in the left ventricular cavity.

CT Pulmonary Vascular Volumetry

CT pulmonary vascular volumetry was performed on a commercially available workstation (Advantage Workstation 4.6; GE Healthcare, Milwaukee, WI, USA). After the bony thorax was removed, an attenuation threshold of 80 HU was applied to isolate the thoracic cardiovascular structures from other soft tissues, and the resulting CT data were saved (Fig. 1A). The CT data corresponding to the central section were extracted from the overall thoracic cardiovascular structure data by eliminating peripheral section data at the pulmonary artery bifurcation, and the atrial-pulmonary venous junctions (Fig. 1B). The right and left pulmonary vessels (Fig. 1C) were segmented by subtraction of the central section from the entire thoracic cardiovascular structures. Stents were excluded from pulmonary vessels in four patients with pulmonary artery stents by applying an attenuation threshold of 800 HU. The segmented right and left pulmonary vessel volumes were calculated with the workstation “volumetry” menu. An approximate post-processing time of 15 minutes was required for each patient.

Fig. 1. 4-year-old girl with repaired left pulmonary artery sling.

A–C. Frontal volume-rendered CT images showing thoracic cardiovascular structures segmented by applying lowest attenuation threshold of 80 HU (A), volume after exclusion of branch pulmonary arteries and pulmonary veins (B), and pulmonary vascular volumes obtained by subtracting central section from entire cardiovascular structure (C). D–I. Additional attenuation thresholds were applied to pulmonary vascular volumes by serially increasing value by 40 HU, up to 600 HU. C–I. Quantified percentage of CT right pulmonary vascular volume was gradually increased from 58.4% at 80 HU to 70.9% at 560 HU. HU = Hounsfield units

Fixed Attenuation Threshold

Attenuation thresholds were serially increased from 80 HU to 600 HU in 40 HU increments to determine the optimal fixed threshold (Fig. 1C–I). We used a wide threshold range based on visually determined attenuation thresholds (120–160 HU) reported in a previous study, to avoid missing the optimal fixed threshold (1).

Adaptive Attenuation Threshold

The average pulmonary artery enhancement (PA_avg) was calculated from the right and left branch pulmonary arterial enhancement values measured by placing a round region of interest in the central half of the branch pulmonary artery. Next, adaptive attenuation thresholds were individually calculated by dividing PA_avg with 2.50 (PA_avg/2.50), 2.00 (PA_avg/2.00), 1.75 (PA_avg/1.75), 1.63 (PA_avg/1.63), 1.50 (PA_avg/1.50), and 1.25 (PA_avg/1.25), respectively.

Lung Perfusion Scintigraphy

After intravenous injection with a body surface area-indexed percentage of the standard 80 MBq adult dose of technetium-99m macroaggregated albumin, lung perfusion scintigraphy was performed with a gamma camera. Pulmonary blood flow was reflected by the distribution of radioactivity in the lungs. Lung fields were viewed in anterior and posterior planes, and the relative perfusion percentage was calculated with geometric mean counts for each lung (Fig. 1C). The right lung perfusion percentage determined by lung perfusion scintigraphy was regarded as the reference standard to determine the optimal attenuation threshold for quantifying the CT right pulmonary vascular volume percentage.

Statistical Analysis

Statistical analyses were performed with Excel (Microsoft Corp., Redmond, WA, USA). Continuous variables were expressed as the mean ± standard deviations or median with range, and categorical variables were expressed as frequency with percentage. The right pulmonary vascular volume percentages measured by cardiothoracic CT, and the right lung perfusion percentages measured by lung perfusion scintigraphy for each threshold were compared and correlated with the paired t test and the Pearson correlation coefficient. A p-value of less than 0.05 was considered statistically significant. In addition, absolute differences between the two values were calculated and averaged. The optimal threshold definition for the CT pulmonary vascular volume ratio included: 1) the smallest mean absolute difference (≤ 5.0%), 2) the lowest statistically significant difference (p-value) compared to the reference values measured by lung perfusion scintigraphy, and 3) the highest Pearson correlation coefficient.

RESULTS

The right lung perfusion percentage value measured with lung perfusion scintigraphy was 69.1 ± 15.0%. The CT right pulmonary vascular volume percentages determined with fixed thresholds were relatively stable in all patients from 80 HU to 400 HU, but began to show substantial fluctuations in some patients from 440 HU to 600 HU (Fig. 2). The mean CT right pulmonary vascular volume percentages determined with the fixed thresholds were initially 65.8% at 80 HU, and gradually increased and peaked to 70.4% at 440 HU (Table 1). The values were significantly lower than the right lung perfusion percentages measured with perfusion scintigraphy in the 80–200 HU range (p < 0.02), but no significant differences were found in the 240–600 HU range (p > 0.05) (Table 1). The mean absolute difference between the two imaging modalities was less than 5.0% in the 160–400 HU range (Table 1). High Pearson correlation coefficients (R = 0.93) were found in the 80–400 HU range, and subsequently gradually decreased at fixed thresholds greater than 440 HU (Table 1). Consequently, 320 HU, 360 HU, and 400 HU fixed attenuation thresholds were regarded as optimal for CT pulmonary vascular volume ratio calculations.

Fig. 2. Graphical illustration of changes in percent of CT right pulmonary vascular volume with fixed attenuation thresholds in 80–600 HU range by 40 HU increments.

Table 1. Comparisons and Correlations between Percentages of CT Right Pulmonary Vascular Volumes Quantified with Fixed Attenuation Thresholds and Nuclear Medicine Right Lung Perfusion Percentages.

Fixed Threshold	80	120	160	200	240	280	320*	360*	400*	440	480	520	560	600
CT right pulmonary vascular volume percentage (%)	65.8 ± 13.8	66.1 ± 14.1	66.5 ± 14.5	67.1 ± 14.9	67.6 ± 15.4	67.9 ± 15.9	68.4 ± 16.5	69.3 ± 17.0	70.0 ± 17.2	70.4 ± 16.9	69.6 ± 18.9	68.7 ± 21.0	69.3 ± 24.6	69.8 ± 26.0
p-value of paired t test	< 0.001	< 0.001	< 0.003	< 0.02	> 0.06	> 0.1	> 0.4	> 0.7	> 0.2	> 0.2	> 0.5	> 0.8	> 0.6	> 0.6
Pearson correlation coefficient	0.93	0.93	0.93	0.93	0.93	0.93	0.93	0.93	0.93	0.90	0.80	0.69	0.63	0.48
Mean absolute difference (%)	5.4	5.3	5.0	4.8	4.6	4.6	4.7	4.6	4.9	5.6	9.1	12.7	17.9	22.1

^*Optimal thresholds for CT pulmonary vascular volume ratio showing smallest mean absolute difference (≤ 5.0%), lowest level of statistically significant difference compared to reference values measured by lung perfusion scintigraphy, and highest Pearson correlation coefficient. HU = Hounsfield units

The mean pulmonary artery enhancement value was 529.5 ± 164.8 HU (range, 250.1–956.6 HU). The mean CT right pulmonary vascular volume percentages determined with adaptive thresholds were initially 67.2% at PA_avg/2.50, and gradually increased and peaked at 71.1% at PA_avg/1.25 (Table 2). Only the values observed at PA_avg/2.50 were significantly lower than the right lung perfusion percentages measured with lung perfusion scintigraphy (p < 0.03), and other values showed no significant differences (p > 0.05) (Table 2). The mean absolute difference between the two imaging modalities was less than 5.0% for all adaptive thresholds, except for PA_avg/1.25 (Table 2). Pearson correlation coefficients were high (R = 0.91–0.92) for all adaptive thresholds (Table 2). Therefore, PA_avg/1.75, PA_avg/1.63, and PA_avg/1.50 were determined as optimal among the six different adaptive thresholds for calculating CT pulmonary vascular volume ratio.

Table 2. Comparisons and Correlations between Percentages of CT Right Pulmonary Vascular Volumes Quantified with Adaptive Attenuation Thresholds and Nuclear Medicine Right Lung Perfusion Percentages.

Adaptive Threshold	PAavg/2.50	PAavg/2.00	PAavg/1.75*	PAavg/1.63*	PAavg/1.50*	PAavg/1.25
CT right pulmonary vascular volume percentage (%)	67.2 ± 14.8	68.0 ± 15.3	68.7 ± 15.9	69.3 ± 16.2	69.7 ± 16.6	71.1 ± 17.2
p-value of paired t test	< 0.03	> 0.2	> 0.6	> 0.8	> 0.5	> 0.05
Pearson correlation coefficient	0.92	0.92	0.91	0.91	0.91	0.91
Mean absolute difference (%)	4.9	4.7	4.8	4.9	5.0	5.5

^*Optimal thresholds for CT pulmonary vascular volume ratio showing smallest mean absolute difference (≤ 5.0%), lowest level of statistically significant difference compared to reference values measured by lung perfusion scintigraphy, and highest Pearson correlation coefficient; PA_avg/2.50 PA_avg divided by 2.50, PA_avg/2.00 PA_avg divided by 2.00, PA_avg/1.75 PA_avg divided by 1.75, PA_avg/1.63 PA_avg divided by 1.63, PA_avg/1.50 PA_avg divided by 1.50, PA_avg/1.25 PA_avg divided by 1.25. PA_avg = average pulmonary artery enhancement

The calculated ratios of CT pulmonary vascular volumes were comparable between optimal fixed and adaptive attenuation thresholds, and were superior to those published in a previous study with visually determined attenuation thresholds (120–160 HU) (5) (Table 3).

Table 3. Comparisons between Optimal Attenuation Thresholds from This Study and Visually Determined Thresholds from Previous Study (5).

Attenuation Threshold	Fixed (320 HU)	Fixed (360 HU)	Fixed (400 HU)	Adaptive (PAavg/1.75)	Adaptive (PAavg/1.63)	Adaptive (PAavg/1.50)	Visually-Determined (120–160 HU) (5)
CT right pulmonary vascular	68.4 ± 16.5	69.3 ± 17.0	70.0 ± 17.2	68.7 ± 15.9	69.3 ± 16.2	69.7 ± 16.6	66.3 ± 14.0
p-value of paired t test	> 0.4	> 0.7	> 0.2	> 0.6	> 0.8	> 0.5	0.001
Pearson correlation	0.93	0.93	0.93	0.91	0.91	0.91	0.92
Mean absolute difference (%)	4.7	4.6	4.9	4.8	4.9	5.0	5.2

PA_avg/1.75 PA_avg divided by 1.75, PA_avg/1.63 PA_avg divided by 1.63, PA_avg/1.50 PA_avg divided by 1.50.

DISCUSSION

In this study, we determined optimal attenuation thresholds for quantifying CT pulmonary vascular volume ratios. Our results demonstrated comparable outcomes for fixed and adaptive optimal thresholds. On average, the optimal adaptive thresholds corresponded to approximately 303 HU for PA_avg/1.75, 325 HU for PA_avg/1.63, and 353 HU for PA_avg/1.50. In fact, the mean attenuation values for optimal adaptive thresholds were in the same range (320–400 HU) for optimal fixed attenuation thresholds. Therefore, higher attenuation thresholds than those visually assessed in a previous study (120–160 HU) (5) should be used to obtain more accurate CT pulmonary vascular volume ratios. We previously theorized that pulmonary microcirculation including capillaries, arterioles, and venules demonstrating relatively lower attenuation values, might contribute significantly to high-fidelity pulmonary vascular volume ratios. However, based on the results of this study, we concluded that excluding enhancing soft tissue around the pulmonary vasculature from pulmonary vascular volume quantification might be crucial for obtaining more accurate ratios of CT pulmonary vascular volumes. Although there were no patients with lung atelectasis showing a high degree of contrast enhancement in this study population, higher attenuation thresholds may exclude atelectasis more effectively from pulmonary vessels.

As previously mentioned, this study demonstrated that the optimal adaptive thresholds proportional to PA_avg provided comparable results to optimal fixed thresholds. Theoretically, the optimal adaptive threshold is presumed superior to the optimal fixed threshold, as demonstrated in a previous study with contrast-enhanced cardiac CT (14). Despite the comparable results, we plan to apply optimal adaptive attenuation thresholds to quantify CT pulmonary vascular volume ratios in future clinical studies, when considering the wide range (250.1–956.6 HU) of PA_avg observed in this study.

There were no significant differences between the percentages of CT right pulmonary vascular volumes quantified with optimal thresholds and nuclear medicine right lung perfusion percentages. It is noteworthy that a small difference may result from the fundamental methodological discrepancy in quantifying the corresponding ratios between cardiothoracic CT and lung perfusion scintigraphy; i.e. pulmonary venous volumes were included in the former but not in the latter. In cyanotic congenital heart disease, the increase in systemic-pulmonary collateral arterial flow—which contributes to the pulmonary venous volumes—tends to be proportional to reduced pulmonary arterial flow (24). The difference between the two methods may be reduced when pulmonary arterial and venous volumes measured with cardiothoracic CT are separately quantified. However, such meticulous separation between pulmonary arteries and veins is one of the most technically challenging segmentation tasks, even with the assistance of a sophisticated algorithm (25).

Notably, cardiothoracic CT can provide not only pulmonary artery morphology but also differential pulmonary vascular volume ratios without additional radiation exposure to the patient. Its clinical usefulness in monitoring the effectiveness of pulmonary artery angioplasty was recently reported (7). The capability to provide CT pulmonary vascular volume ratio may expand the clinical role of cardiothoracic CT, particularly with respect to functional or quantitative evaluations, such as ventricular volumetry (26), and myocardial delayed enhancement (27). Additionally, the CT pulmonary vascular volume ratio could become a fundamental thoracic imaging method for assessing the functional aspects of various pulmonary diseases (28). For example, a central-to-peripheral pulmonary vascular volume ratio measured with CT may be used to evaluate the presence and severity of pulmonary arterial hypertension.

This study has several limitations. First, cardiothoracic CT and lung perfusion scintigraphy were performed on different days in most patients. Therefore, a change in right-to-left pulmonary vascularity ratio might have occurred during the elapsed time interval. The limitation stemming from the retrospective design of this study warrants a future prospective study. Second, a statistical comparison was not carried out to determine the optimal attenuation threshold in this study, since the differences between different thresholds were not enough to predict statistical significance. Therefore, optimal attenuation thresholds were descriptively defined to have the highest correlation, no statistically significant difference, and a mean absolute difference ≤ 5.0%, with reference values measured using lung perfusion scintigraphy.

In conclusion, the optimal fixed and adaptive thresholds for quantifying CT pulmonary vascular volume ratios appear equally useful. Considering the wide range of PA_avg, the application of optimal adaptive, rather than fixed thresholds may be more suitable in actual clinical practice.