A direct comparison of 3 T contrast-enhanced whole-heart coronary cardiovascular magnetic resonance angiography to dual-source computed tomography angiography for detection of coronary artery stenosis: a single-center experience

This prospective study was approved by the local ethics committee; all patients gave written informed consent. From June 2014 until January 2015, 74 patients with known or suspected CAD were enrolled. The patients were scheduled for conventional x-ray coronary angiography. The exclusion criteria were: absence of a sinus cardiac rhythm, orthopnea, history of previous heart surgery, presence of coronary stent, and contraindications to CMR imaging (claustrophobia, pacemaker) or CT with known allergy to iodinated contrast material, impaired renal function (creatinine > 1.4 mg/dl) and thyroid disorders.

Patients with heart rates> 75 bpm received an oral β-blocker (metoprolol tartrate, 25–50 mg) before CE-CCMRA. In addition, 0.5 mg nitroglycerin was administered sublingually to all subjects before CE-CCMRA. A medical abdominal belt was wrapped tightly along the side of the ribs to reduce abdominal motion during deep inspiration. No β-blocker or nitroglycerin was administered to patients before DS-CCTA scans.

CE-CCMRA was performed using a 3 T whole-body scanner (MAGNETOM Trio; Siemens Healthineers, Erlangen, Germany) with a 12-channel matrix coil (6 each for dorsal and ventral, respectively). The R-wave acquired from a three-lead wireless vectorcardiogram was used to trigger the data acquisition. The CE-CCMRA scanning procedure was as follows: after initial localization of 2-chamber and 4-chamber views and left ventricular short-axis views, a retrospective electrocardiography (ECG)-triggered cine (a fast low-angle shot (FLASH) sequence with 80 cardiac phases) was performed in the 4-chamber plane to determine the quiescent period for coronary artery imaging [8]. The acquired cine movie was visually assessed to calculate the patient-specific trigger-delay time and duration of data acquisition. For whole-heart CE-CCMRA, a prospective navigator-gated, ECG-triggered, fat-saturated, inversion recovery-prepared, segmented, three-dimensional FLASH sequence (TR 320 ms, TE 1.4 ms; flip angle 20°; matrix 256 × 256, FOV 220 × 330 mm², acquired voxel size = 1.3 × 1.3 × 1.8 mm³ and interpolated to 0.65 × 0.65× 0.9 mm³) was employed. The three-dimensional k-space data were collected with a centric ordering scheme in the phase-encoding direction and a linear order scheme in the partition-encoding direction. In addition, a nonselective inversion pulse with an inversion time (TI) of 200 ms was applied before the navigator echo pulses to suppress the background tissues. In order to reduce the acquisition time, parallel imaging technique (generalized auto-calibrating partially parallel acquisitions, GRAPPA) was used in the phase-encoding direction with an acceleration factor of 2. A 0.2 mmol/kg intravenous injection of contrast agent (Gadobenate dimeglumine, Multi-Hance; Bracco Imaging SpA, Milan, Italy) was injected into an antecubital vein at a rate of 0.3 ml/s, immediately followed by 20 ml saline given at the same rate. Data acquisition began 60s after the initialization of contrast agent administration.

All patients underwent DS-CCTA (SOMATOM Definition, Siemens Healthineers). The DS-CCTA protocol included the following parameters: retrospective ECG gating; collimation, 32 × 0.6 mm; slice acquisition, 64 × 0.6 mm using the z-flying focal spot technique; gantry rotation time, 330 ms; pitch, 0.20–0.43 adapted to the heart rate; tube voltage, 100 or 120 kVp (depending on age and body mass index); and maximum tube current, 320 mAs per rotation. Spatial resolution was 0.6 × 0.6 × 0.6 mm. Temporal resolution was variable according to the heart rate of the patient.

All DS-CCTA data were acquired in deep inspiration. The volume of iodinated contrast material (Iopamidol 370 mg of iodine/ml, Bracco) was adapted to the scan time. A contrast bolus (60 to 90 ml adjusted to the individual’s body weight and scan time) was injected in an antecubital vein with a flow rate of 4 to 6 ml/s followed by a saline chaser of 40 ml at the same rate. Contrast agent application was controlled by a bolus tracking technique. The region of interest was placed into the aortic root, and image acquisition was started 7 s after the signal density level reached the predefined threshold of 100 HU.

All DS-CCTA coronary angiography datasets were reconstructed with a slice thickness of 0.75 mm and increment of 0.4 mm using medium-to-smooth convolution kernel (B26f), resulting in an inplane resolution of 0.6 to 0.7 mm and a through-plane resolution of 0.5 mm [9].

Conventional x-ray coronary angiography served as the reference standard in this study. The diameter stenosis, as a percentage of the reference diameter, was determined in two orthogonal directions and the measurement was performed in the projection that showed the highest degree of stenosis, and was evaluated by two experienced invasive cardiologists without knowledge of the results of CE-CCMRA and DS-CCTA. All segments of the coronary artery tree with a reference diameter of 1.5 mm or greater were included in the study. A clinically significant stenosis was defined as stenosis of at least 50% of the vessel diameter.

Transverse images were transferred to an off-line workstation (Siemens Healthineers). Images for each patient were analyzed in consensus by two cardiovascular radiologists with 16 and 20 years of experience. Two observers blind to the results of conventional coronary angiography evaluated the DS-CCTA and CE-CCMRA results using axial slices images, as well as multiplanar or curved reformatted reconstructions and maximum intensity projections. A 16-coronary-arterysegment model was used, based on recommendations from the American Heart Association (modified 15-segment model, with segment 16 being the intermediate branch of the left coronary artery). Image quality was assessed on a 4-point scale: 1 = poor (not assessable); 2 = fair (mild to moderate artifacts); 3 = good (minimum to mild artifacts); and 4 = excellent (minimum or no artifacts). Contrast-to-noise ratio (CNR) was evaluated in nondiseased proximal coronary segments (proximal left main (LM), mid-left anterior descending coronary artery (LAD), left circumflex coronary artery (LCX), and right coronary artery (RCA)) and computed as the difference in mean signal intensity between the vessel lumen and the adjacent tissue, divided by the standard deviation of the background noise in the proximal aorta [10]. The severity of luminal diameter reduction as being < 50% or ≥ 50% was visually assessed by two readers independently.

The evaluation of segments by conventional x-ray coronary angiography, 3 T CE-CCMRA and DS-CCTA for 51 patients are shown in Table 2. A total of 76 of 654 coronary segments (12%) with a reference luminal diameter ≥ 1.5 mm by x-ray angiography were not available for evaluation by CE-CCMRA. However, only 35 of the 654 coronary segments (5%) with a reference luminal diameter ≥ 1.5 mm by x-ray angiography could not be visualized by DS-CCTA due to poor CNR, motion artifacts, and small diameter. For all vessels, DS-CCTA demonstrated higher image quality than CE-CCMRA. In addition, the subjectively graded image quality of segments imaged by DS-CCTA (3.5 ± 0.8) was judged to be higher than that of CE-CCMRA (3.2 ± 0.7, p < 0.001). The kappa values for interobserver agreement for image-quality grading were 0.86 and 0.83, respectively. Representative examples of normal coronary angiograms and coronary stenoses detected by CE-CCMRA and DS-CCTA are shown in Figs. 2 and 3, respectively.

The diagnostic abilities of CE-CCMRA and DS-CCTA to detect significant stenoses in a patient-, and vessel-based analysis are detailed in Table 3 and Fig. 4.

In the patient-based analysis, DS-CCTA correctly recognized 29 of the 31 significant stenoses detected on x-ray angiography. CE-CCMRA also correctly identified significant CAD in 29 out of 31 patients. The sensitivity (93.5%) of both tests was equally high. In contrast, CE-CCMRA had a slightly lower specificity (85%) but no significant difference in the ability to detect coronary stenosis than DS-CCTA (90%). CE-CCMRA had a AUC of 0.89 (95% CI: 0.79 to 0.99). DS-CCTA had a AUC of 0.92 (95% CI: 0.83 to 1.00). In the vessel -based analysis, the sensitivity, specificity, PPV, and NPV for detecting patients with significant CAD by DS-CCTA scan were 94.7%(54/57), 90.7%(87/96), 85.7%(54/63), and 96.7%(87/90), respectively. In addition, the diagnostic performance of CE-CCMRA on a per-vessel basis was similar, with no significant differences. The AUC was 0.91 (95% CI: 0.86 to 0.97) for CE-CCMRAand 0.93 (95% CI: 0.88 to 0.98) for DS-CCTA.

Differences were observed for the RCA, LM-LAD, and LCX arteries (Table 4, Fig. 5). The sensitivity, specificity, positive predictive valve (PPV), negative predictive value (NPV) of the left main–LAD, RCA, LCX were similar with CE-CMRA and DS-CCTA with no significant differences. The AUC of RCA, LM-LAD, and LCX was 0.94, 0.92, 0.86 for CE-CCMRA and 0.93, 0.94, 0.91 for DS-CCTA, respectively. However, for the LCX, both CE-CCMRA and DS-CCTA had low positive predictive values.

This study also has some limitations. First, the study population from the single center was relatively small. We will continue our studies with a large patient population for further subgroup analysis. Another drawback is that, after contrast administration, DS-CCTAand CE-CCMRA are not suitable for repeat examinations when the scans fail. Moreover, the time for completing CE-CCMRA procedures still remains long, and CE-CCMRA has relatively low spatial resolution. Third, the evaluation of the diagnostic performance of 3 T CE-CCMRA in this study was performed using a 12-channel cardiac coil. Recently, a 32-channel cardiac coil has become commercially available, which would reduce the imaging time of cardiac imaging [30], and could offer even higher diagnostic accuracy [31]. Finally, a limitation of this approach is that it may be difficult to combine a high dose CE-CCMRA scan with a rest and stress perfusion CMR protocol which would require additional gadolinium.