Dual-source computed tomography in patients with acute chest pain: feasibility and image quality

Sixty consecutive patients (32 females, 28 males, mean age 58.1±16.3 years, age range 26–84 years) were prospectively included in this study. Intake was performed on weekdays from 7 am to 7 pm from August to October 2006. All patients suffered from acute chest pain and were referred to our department to diagnose or to rule out pulmonary embolism (n=56) or aortic dissection (n=4). Inclusion criteria were acute chest pain >5 min within the previous 24 h and/or elevated serum D-dimer levels. Dyspnea and hemodynamic instability were not considered exclusion criteria. Similarly, all patients irrespective of their mean or regularity of heart rate and irrespective of their ability to perform breath-hold were included. Exclusion criteria included pregnancy, previous adverse reaction to iodinated contrast agent, nephropathy (serum creatinine >1.3 mg/dl), elevated troponine-I or creatine kinase-MB level in the initial blood sample, initial diagnostic ECG changes indicating an acute coronary syndrome (i.e., ST elevation or depression >1 mm, T-wave inversion >4 mm in >2 anatomically contagious leads), and interference with standard clinical care of patients. The study was approved by the local ethics committee; informed consent was obtained.

All CT examinations were performed on a DSCT scanner (Somatom Definition, Siemens Medical Solutions, Forchheim, Germany). First, a single non-enhanced low-dose scan at the level of the aortic root was obtained. In this slice, a region of interest (ROI) was set in the lumen of the aorta for monitoring intraluminal contrast enhancement. The delay from start of contrast material injection to start of scanning was planned using the bolus-tracking technique. A total of 110 ml iodinated contrast material (iodixanol, Visipaque 320; 320 mg/ml, GE Healthcare, Buckinghamshire, UK) was administered at a flow rate of 4 ml/s via an 18-gauge needle placed into a superficial vein in the left antecubital fossa, followed by 30 ml saline solution at the same flow rate (4 ml/s). After reaching the preset contrast enhancement level of 80 Hounsfield Units (HU) in the ROI, a breath-hold signal was given and the scan was initiated automatically after a delay of 6 s. A topogram was used for planning the examination and determining the scan range. Data acquisition was performed using a dedicated biphasic chest pain protocol. As illustrated in Fig. 1, the lower chest including the heart was scanned with a tube current time of 320 mAs while the upper chest was scanned with a tube current time product of 160 mAs. The borderline between upper and lower scan range was set at approximately 2 cm below the carina. Data acquisition was performed in a cranio-caudal direction. Acquisition parameters were as follows: detector collimation 2×32×0.6 mm by using a z-flying focal spot for the simultaneous acquisition of 2×64 overlapping 0.6-mm slices, gantry rotation time 330 ms, and tube potential 120 kV. The pitch varied according to the patient's heart rate and ranged from 0.2–0.43, with higher pitch at higher heart rates. ECG-pulsing for radiation dose reduction [22] was applied in all patients. At mean heart rates below 60 bpm, full tube current was applied from 60 to 70%, at 61–70 bpm from 50 to 80%, and at heart rates above 70 from 30 to 80% of the R-R interval.

Retrospective ECG-gating for phase synchronization was used. For the heart, CT data sets were reconstructed at 70% of the R-R interval with a slice thickness of 0.75 mm (increment 0.5 mm) by using a medium soft-tissue convolution kernel (B26f) (mean field of view, FoV: 151±17 mm, image matrix 512×512). If considered necessary, additional images were reconstructed in 5% steps using the same parameters within the time window of full tube current. Images of the mediastinum (mean FoV: 293±43 mm) including the aorta and pulmonary arteries were reconstructed with a slice thickness of 1 mm (increment 0.8 mm) by using a medium soft-tissue convolution kernel (B30f), and images of the lung were reconstructed with a slice thickness of 2 mm (increment 1.5 mm) by using a sharp convolution kernel (B60f, same FoV as for the mediastinum). All images were transferred to a second Wizard (Siemens) equipped with cardiac post-processing software (Syngo Circulation, Siemens).

All data were qualitatively evaluated regarding image quality and artifacts of different thoracic structures by two independent readers who are both experienced in cardiovascular radiology. This evaluation was performed on transverse source images, multi-planar reformations (MPR), curved MPR, and thin-slab maximum intensity projections.

Image quality of lung parenchyma was independently rated using a two-point scale, adapted from a previous publication [23]. Lung parenchyma allowing diagnostic assessment due to distinct anatomic details of bronchial and parenchymal structures without significant artifacts and noise was rated with a score of 1 (diagnostic). Lung parenchyma with artifacts or noise causing reduction of image quality and diagnostic value was rated with a score of 2 (non-diagnostic).

Artifacts were rated to quantify the cranio-caudal distribution of artifacts within the lung parenchyma for the right and left lung separately in coronal MPR. Because of the different tube current at upper and lower parts along the z-axis (see Fig. 1), the following scores were separately applied for the apex and the basis of the lung, respectively: 1= no artifacts, 2= breathing artifacts (stair step artifacts), 3= ECG-gating (i.e., synchronization or interpolation artifacts), and 4= noise artifacts. If breathing and ECG-gating artifacts appeared at the same time in one patient, the artifact with the worst impact on image quality was noted.

Image quality of thoracic vascular structures was independently rated using the same two-point scale [23] as used for the rating of the lung parenchyma. Regarding the thoracic aorta, a score of 1 (diagnostic) indicated confident evaluation of the ascending aorta, the aortic arch, and descending aorta. Regarding pulmonary arteries, a score of 1 indicated confident evaluation of central, lobar, segmental, or subsegmental pulmonary arteries. Regarding coronary arteries, a score of 1 represented confident depiction (homogenous attenuation; no artifacts decreasing coronary analysis) of the right coronary artery (RCA), left main artery (LMA), left anterior descending artery (LAD), left circumflex artery (LCX), and their side branches. A score of 2 indicated decreased image quality of thoracic vascular structures with severe impairment of diagnostic value (non-diagnostic) due to breathing, motion, or ECG-gating artifacts.

Image noise was determined as the standard deviation of attenuation in a ROI placed in the ascending aorta [24]. Contrast attenuation was measured in each patient in the ascending aorta, pulmonary trunk, LMA, proximal segment of the RCA, and right and left ventricle. Image noise and contrast attenuation were assessed using a circular ROI positioned exactly within the vessel or ventricular lumen while avoiding superimposition or partial volume effects from the vessel wall or myocardium. Measurements of the ascending aorta and pulmonary trunk as well as of the right and left ventricle were performed on the same transverse image, while the ROI in the coronary arteries was individually placed on separate transverse images.

Imaging findings indicating the possible underlying cause of acute chest pain were documented in each patient by both readers in consensus.

Statistical analysis was performed using commercially available software (SPSS 11.5, SPSS Inc., Il). Quantitative variables are expressed as mean ± standard deviation (SD) including 95% confidence intervals (CI) or range when appropriate. Categorical data were expressed as frequencies or percentages. Inter-observer agreement (kappa statistics) for image quality ratings was calculated. Two-tailed Student's t test for paired samples was used to explore significant differences in vessel attenuation among the ascending aorta, RCA, LMA, pulmonary trunk, and right and left ventricle. Intra-individual differences regarding vessel attenuation among the different vascular territories were performed using pairwise comparisons. Bonferroni correction for multiple comparisons was made, and a P-value <0.003 was considered statistically significant.