All patients who met the inclusion criteria agreed to participate in the study and signed an informed consent form. The study protocol was approved by the Institutional Review Board from our hospital.

Patients with suspected or known coronary artery disease were prospectively enrolled from August 2016 to December 2017. For cases involving routine CCTA, exclusion criteria were a history of allergic reaction to iodinated contrast agents, arrhythmia, previous cardiac surgery, evidence of acute coronary syndrome, New York Heart Association functional class III or IV heart failure, aortic stenosis, and intolerance to beta-blockers.

For subtraction CCTA scans, the inclusion criteria were heart rate (HR) ≤ 65 beats per minute (bpm) after oral administration of beta-blocker (Betaloc, 25 mg; AstraZeneca, London, UK) or basal HR ≤ 65 bpm and coronary artery calcification detected in the proximal segments. Proximal coronary segments were defined as previously described (9). Patients were excluded for further mask scans due to motion artifacts in contrast images.

During the study period, a total of 164 patients underwent subtraction CCTA due to coronary artery calcification or stent placement. Of these 164 patients, 36 patients (33 with coronary stents, 2 with motion artifacts, and 1 with insufficient coronary lumen contrast enhancement) were excluded from the study. Finally, 128 patients who had undergone subtraction CCTA scanning were included for further analysis, of which 32 (25%) underwent CCTA with invasive coronary angiography (ICA) as the reference standard to determine the diagnostic value.

ICA using the standard technique was performed in clinically indicated subjects within 60 days of subtraction CCTA scanning. ICA (Anigostar Plus; Siemens Healthineers, Berlin, Germany) was performed using standardized angiographic techniques. Quantitative coronary analysis (QCA) was performed using a software program (Syngo fastView, VX57I33; Siemens Healthineers) by a cardiologist with more than 15 years of experience. Significant stenosis was defined as > 50% luminal reduction in the most severe stenosis in the calcified segments.

All scans were performed using a 320-detector row CT scanner (Aquilion ONE VISION Edition; Canon Medical Systems, Otawara, Japan) with 0.5-mm detector elements and a gantry rotation time of 275 ms. Images were reconstructed with a 512 × 512 matrix and 0.5-mm thickness by using standard kernel FC09 iterative reconstruction (AIDR 3D; Canon Medical Systems). Beta-blockers were given when the resting HR was above 65 bpm unless patients had contraindications. Sublingual short-acting nitroglycerin (0.5 mg; Yimin Pharma, Beijing, China) was routinely given during data acquisition.

Noncontrast CT scanning was performed for all patients to determine the calcification score using the method described by Agatston et al. (16). Noncontrast cardiac CT scans were obtained with a tube voltage of 120 kV and a tube current of 50 mA. Mid-diastolic or end-systolic images were acquired using the prospective gating method. The scan range included the region from the origin of the aorta to the apex of the heart. The reconstruction slice thickness was 3 mm.

We adopted a modified two-breathhold approach in this study (Fig. 1). After the contrast imaging acquisition, the investigator determined whether the target coronary segments needed the mask images for postprocessing subtraction. It took at least 60 seconds after the contrast injection until the next noncontrast scan. Theoretically, there was no contrast residue in the coronary artery system and left cardiac chamber during mask scanning. Both the contrast and mask image data were acquired by using a prospective electrocardiography triggering technique between 65% and 80% of the RR interval. Scanning was performed with a tube voltage of 120 kVp and a rotation time of 275 ms. The tube current setting was selected based on the value calculated by the automatic exposure control function (17) with the standard deviation (SD) value set to 20 HU.

The contrast-enhanced scan was performed using iodinated contrast medium with an iodine concentration of 370 mg/mL (Ultravist® iopromide-370; Bayer AG, Berlin, Germany) injected at a rate of body weight (kg) × 0.06 mL/s in 10 s (fixed), followed by a 30-mL saline chaser bolus. The effective radiation dose was estimated based on the dose-length product (DLP, mGy × cm) by using the formula of effective radiation dose = DLP × κ, where κ = 0.014 mSv × mGy⁻¹ × cm⁻¹ (18).

Subtraction was performed using a dedicated algorithm “volumetric CT digital subtraction angiography” (19) that was available on the scanner console. First, a global nonrigid registration was performed, followed by a local rigid registration for the targeted coronary segment. Thus, the results obtained reflected only the contrast filling the coronary artery without calcification.

The acquired conventional and subtraction CCTA images were transferred to a workstation (Vitrea, version 6.5.3; Vital, Minneapolis, MN, USA) to generate a curved planar reformation image and a cross-sectional image. The coronary arteries were divided into 17 segments according to the American Heart Association classification. Plaque calcification within each coronary segment was determined as follows: in addition to the most severe plaque calcification identified by CCTA using a cross-sectional grading method as previously described (20), we assessed the plaque calcification associated with the maximum cross-sectional arc within each segment as defined by CCTA for this investigation. To analyze the stenosis in the targeted segment, the minimum lumen diameter was measured at the site with the highest degree of arc calcifications within each segment. On the basis of the extent of segment calcification determined using an adapted cross-sectional arc calcification method (21), the targeted calcified plaques were classified as mild calcification (cross-sectional arc calcium < 90°), moderate calcification (cross-sectional calcium arc 90°–180°), and severe calcification (cross-sectional calcium arc > 180°) (Fig. 2).

Segmental image quality was evaluated for each calcified segment using a 4-point scale: 1 = uninterpretable, 2 = poor image quality, 3 = adequate image quality, 4 = good image quality. Segments with two points due to distinct misregistration were deemed nondiagnostic and were considered to be equivalent to severe stenosis for subsequent analysis.

Quantitative assessment of the stenosis was performed in cross-sectional and longitudinal projections by two experienced cardiac radiologists with more than 10 years of experience. The radiologists were blinded to the results of ICA but were aware of the clinical history of the patients. Stenosis was graded as minimal, mild, moderate, and severe, corresponding respectively to stenosis of ≤ 25%, 26–50%, 51–75%, and ≥ 76%, as recommended by the Society of Cardiovascular Computed Tomography (22). Discrepancies were settled by consensus. Note that if the targeted segment was judged to be non-assessable, the image findings were classified as severe stenosis.

We evaluated the data on a segment level. The average image quality scores between conventional and subtraction CCTA images were compared using a paired t test. The diagnostic vs. nondiagnostic image quality frequencies were compared using the McNemar test. The kappa coefficient was used to measure agreement between conventional and subtraction CCTA for targeted segment stenosis grading. For all the ICA examinations performed, the conventional and subtraction CCTA diagnostic accuracies (sensitivity, specificity, positive predictive value, and negative predictive value) for detecting significant stenosis (> 50%) were assessed in segments with severe calcification, and p values for comparison between diagnostic tests were calculated by the McNemar test. To assess diagnostic accuracy, the area under the curve (AUC) of the receiver operating characteristic (ROC) was calculated for both conventional CCTA and subtraction CCTA. The method proposed by DeLong et al. (23) was used for comparison of the two ROC curves. Statistical analyses were performed using the Statistical software MedCalc version 11.5.0.0 (MedCalc Software, Ostend, Belgium). Metric variables were expressed as the mean ± SD. All tests were two-tailed with a significance threshold of p values less than 0.05 and a confidence interval (CI) of 95%.

A total of 128 patients with 564 calcified coronary segments underwent subtraction scanning. With regard to the global calcified segments, the mean coronary image quality score with conventional CCTA was 3.30 ± 0.86 and 3.56 ± 0.62 with subtraction CCTA (p < 0.01). On the basis of conventional CCTA images, 106 (82.8%) patients with 216 (38.3%) segments were judged to have mild calcification, and 94 (73.4%) patients with 185 (32.8%) segments were judged to have moderate calcification. However, image quality in subtraction CCTA did not change significantly in segments with mild calcification (3.68 ± 0.54 to 3.71 ± 0.49, p = 0.68) and segments with moderate calcification (3.49 ± 0.65 to 3.56 ± 0.59, p = 0.48) (Fig. 3). Eighty-four of the 128 subjects were judged to have severe calcification, with severe calcification observed in 163 (28.9%) of the total segments. In the segments showing severe calcifications, after the subtraction of beam-hardening artifacts, the image quality significantly improved from 2.51 ± 0.98 to 3.12 ± 0.94 (p < 0.001) (Fig. 3). Misregistration artifacts were recorded in 17 segments (8%) in the group with mild calcification, 24 segments (13%) in the group with moderate calcification, and 42 segments (26%) in the group with severe calcification.

For the stenosis grading in conventional CCTA and subtraction CCTA, the kappa index was 0.88 (95% CI, 0.80–0.95; p < 0.001) in the group with mild calcification and 0.84 (95% CI, 0.75–0.92; p < 0.001) in the group with moderate calcification. We observed good agreement values between the mild and moderate calcification groups. Nevertheless, for segments with severe calcification, there was a fair correlation between conventional CCTA and subtraction CCTA for the assessment of stenotic grading, with a kappa index of 0.57 (95% CI, 0.44–0.70; p < 0.001). Of the 163 segments with severe calcification, 52 (32.0%) were changed to a different stenotic class after subtraction, among which 21 (40.4%) were downgraded from moderate to mild stenosis, 17 (32.7%) were downgraded from severe to moderate stenosis, 8 (15.4%) were downgraded from severe to mild stenosis, and 6 (11.5%) were upgraded from mild to moderate stenosis. Segments showing grade changes in the mild and moderate calcification groups were less frequent at 11.1% (24/216) and 16.2% (30/185), respectively. Figure 4 shows the cases of stenosis regrading after subtraction.

A total of 25% (32/128) of patients with 58 severely calcified lesions underwent subtraction CCTA scans followed by ICA. In the group with severe calcification, nondiagnostic segments with conventional CCTA were obtained in 33% (19/58) of the cases, and after subtraction, the corresponding percentage was 10% (6/58) (p < 0.01). Statistical performance of conventional CCTA and subtraction CCTA to detect coronary artery stenosis > 50% is shown in Table 2. The AUC for the diagnosis of significant stenosis as determined by QCA (Fig. 5) was 0.94 (95% CI, 0.86–1.00) in subtraction CCTA, which was higher (95% CI, 0.74–0.96) than that for conventional CCTA (0.85; p = 0.03). An example is shown in Figure 6.