Accuracy of Ultrahigh-Resolution Photon-counting CT for Detecting Coronary Artery Disease in a High-Risk Population

Participant Population

Patients with severe aortic valve stenosis and referral for pre-TAVR work-up were enrolled in this prospective study between August 2022 and February 2023. Inclusion criteria were (a) severe aortic valve stenosis confirmed with transthoracic echocardiography, (b) no contraindication to iodinated contrast-enhanced CT, and (c) diagnostic ICA performed within 30 days of CT. Participants who did not meet the predefined inclusion criteria or who did not provide consent were excluded. The institutional review board approved the study protocol (approval date, September 21, 2021; approval number, 21–2469). Informed written consent was obtained.

Photon-counting CT Protocol

All participants were scanned with a first-generation dual-source photon-counting CT scanner (NAEOTOM Alpha, software version syngo CT VA50; Siemens Healthineers). The protocol consisted of (a) an unenhanced coronary calcium scoring scan (14) followed by (b) contrast-enhanced ECG-gated UHR CCTA, and (c) aortoiliac CT angiography. In accordance with consensus statement guidelines for TAVR CT, neither β-blockers nor nitroglycerin were administered to avoid a sudden decrease in blood pressure (15). A dual-syringe power injector was used to administer 70 mL of iopromide (370 mg iodine per milliliter, Ultravist; Bayer Healthcare) followed by a solution of 30 mL iopromide and 40 mL isotonic saline at a flow rate of 5.0 mL/sec. UHR CCTA was initiated via bolus tracking. Scanning parameters for UHR CCTA were as follows: retrospective ECG-gated dual-source helical scan with a pitch factor of 0.2, collimation of 120 mm × 0.2 mm, gantry rotation time of 0.25 second, tube voltage of 120 kV or 140 kV, automated tube current modulation, and dose modulation with ECG pulsing set at 20%–80% of the R-R interval. Automatically determined best systolic, best diastolic, and multiphase data were reconstructed using a moderately sharp vascular convolution kernel (Bv56, quantum iterative reconstruction level 3), a section thickness of 0.2 mm, and an increment of 0.1 mm. The matrix size was set at 1024 × 1024 pixels, and a field of view of 180 × 180 mm was applied.

Subjective and Objective Evaluation of Image Quality

Subjective assessment.—Two radiologists with 4 (M.T.H., in training) and 15 (T.K.) years of experience in cardiovascular imaging assessed the image quality of the UHR CCTA data on a dedicated workstation (syngo.via, version VB60; Siemens Healthineers) in consensus using a five-point Likert scale, as follows: 1, excellent (absence of artifacts); 2, good (mild artifacts); 3, fair (moderate artifacts); 4, poor (pronounced artifacts or low contrast); and 5, nondiagnostic (severe artifacts).

Objective assessment.—The contrast-to-noise ratio and the signal-to-noise ratio per vessel were calculated for the left anterior descending, left circumflex, and right coronary arteries (Appendix S1, Fig S1).

Assessment of Coronary Artery Stenosis in UHR CCTA

The same two readers performed stenosis quantification independently, and both were blinded to the ICA results. Coronary segments at least 1.5 mm in diameter were analyzed. The stenosis diameter was obtained (Appendix S1), and CAD was defined as stenosis of 50% or greater. Segments scored as nondiagnostic due to impaired image quality on every reconstructed CT image were rated as potential stenosis (Fig S2).

ICA as Reference Standard Measurement of Coronary Vessel Stenosis

ICA was performed with standard techniques by board-certified interventional cardiologists (C.v.z.M.), with access gained via the radial or femoral common artery. At least two different projections of each vessel were obtained. For this study, the results of the validated ICA reports were used. No dedicated study read was performed.

Assessment of Radiation Dose

The effective tube current, CT dose index volume, and dose-length product were extracted from the participant protocols. The effective dose was calculated by multiplying the dose-length product by a conversion factor of 0.014 mSv/mGy ∙ cm (16).

Statistical Analyses

SPSS Statistics, version 29.0 (IBM) and MedCalc, version 20.1 (MedCalc Software) were used for statistical analysis. A one-sample Shapiro-Wilk test was used to check the assumption of normal distribution. Categorical variables were expressed as number of findings, with percentages in parentheses. Quantitative variables were expressed as means ± SDs or medians and IQR ranges depending on data distribution. Interobserver agreement for participant-based CT analysis of CAD was evaluated using the Cohen κ statistic (17). To evaluate the diagnostic performance of UHR CCTA compared with the reference standard of ICA, receiver operating characteristic curve analysis was performed, and the area under the receiver operating characteristic curve (AUC) was calculated (18). A nonparametric distribution was assumed for estimating the AUC standard error approximation, and a 95% asymptotic CI was calculated. Sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were calculated for stenosis of 50% or greater at UHR CCTA and stenosis of 50% or greater as well as 70% or greater at ICA on a participant-, vessel-, and segment-based level, and results with 95% CI are reported (19). Furthermore, diagnostic accuracy was evaluated in participant subgroups based on Agatston score (≥1000 or not), stent status, and known or unknown CAD history. Two-tailed P < .05 indicated a significant difference for all tests. Two reviewers (M.S., M.T.H.; 8 and 4 years of experience, respectively) performed all tests.

Patient Characteristics

Of the 72 participants screened, four were excluded from the study because ICA was not performed within 30 days of CT (n = 3) or because the CT scan was aborted (n = 1). Thus, 68 participants (mean age, 81 years ± 7; 32 male, 36 female) with severe aortic valve stenosis and indication for TAVR planning work-up were ultimately included (Fig 1). UHR CCTA was successfully performed in all subjects, and example images are given in Figures 2 and 3. The Agatston scores ranged from 0 to 4273 (median score, 414; IQR range, 125–1246). In the final cohort, 24 of 68 (35%) participants had an Agatston score of 1000 or greater, 18 of 68 (27%) had a known history of CAD, and coronary stents were present in 15 of 68 (22%) participants. Participant characteristics are summarized in Table 1.

Figure 1:

Download as PowerPoint

Figure 2:

Download as PowerPoint

Figure 3:

Download as PowerPoint

Table 1: Participant Characteristics

Subjective and Objective Evaluation of Image Quality

In total, 55 segments were excluded from analysis because of minimal size or because of their absence resulting from anatomic variation. From the 68 participants, 965 analyzed coronary segments were evaluated. Of those, 97.8% (944 of 965) of segments were rated as diagnostic (Likert score, 1–4), whereas 2.2% (21 of 965) were rated as nondiagnostic (Likert score, 5). The median image quality score was 1.5 (IQR range, 1.3–2.0), and most segments were rated as having good (26.0% [251 of 965]) or excellent (52.8% [510 of 965]) quality.

The mean overall contrast-to-noise ratio was 10.5 ± 2.7, and the mean overall signal-to-noise ratio was 9.4 ± 2.6, while the noise level of UHR CCTA at the level of the aortic root of the left coronary artery ostium was elevated (mean noise level, 47.9 HU ± 10.8) (Table 2).

Table 2: Image Quality of Ultrahigh-Resolution CT

Diagnostic Performance of UHR CCTA in Coronary Stenosis Compared with ICA

ICA was used to identify stenosis with a diameter of 50% or greater in 24 of 68 (35%) participants, in 35 of 204 (17.2%) vessels, and in 43 of 965 (4.5%) segments, while 44 of 68 (65%) participants, 169 of 204 (82.8%) vessels, and 922 of 965 (95.5%) segments did not show coronary stenosis. The AUC for detection of CAD, defined as stenosis of 50% or greater, was 0.93 (95% CI: 0.86, 0.99) at the participant level, 0.94 (95% CI: 0.91, 0.98) at the vessel level, and 0.92 (95% CI: 0.87, 0.97) at the segment level (Fig 4).

Figure 4:

Download as PowerPoint

The overall diagnostic accuracy of UHR CCTA in the detection of stenosis of 50% or greater for participant-based, vessel-based, and segment-based analysis was 88% (60 of 68), 91% (185 of 204), and 95% (912 of 965), respectively. Positive and negative prediction values in the detection of stenosis of 50% or greater were 77% (23 of 30) and 97% (37 of 38), respectively. While segment-based, vessel-based, and participant-based specificities in the detection of stenosis of 50% or greater were high (95% [879 of 922], 91% [154 of 169], and 84% [37 of 44], respectively), a decrease in the segment-based sensitivity was observed: 96% (23 of 24) per participant, 89% (31 of 35) per vessel, and 77% (33 of 43) per segment.

ICA further identified stenosis of 70% or greater in 18 of 68 (26%) participants, 27 of 204 (13%) vessels, and 29 of 965 (3%) segments. For detection of stenosis of at least 70%, UHR CCTA showed an accuracy of 82% (56 of 68) per participant, 89% (181 of 204) per vessel, and 94% (906 of 965) per segment. Positive and negative prediction values in the detection of stenosis of at least 70% were 60% (18 of 30) and 100% (38 of 38), respectively. Detailed diagnostic metrics are provided in Table 3. Interobserver agreement for diagnosis of CAD at the participant level was substantial (κ = 0.76, P < .001). For images of nondiagnostic quality, stenosis was assumed in 21 of 965 (2.2%) coronary segments distributed among eight of 204 (3.9%) vessels and seven of 68 (10%) subjects. When compared with ICA results, 18 of 21 (86%) of these segments were falsely suspected of having stenosis, while three of 21 (14%) resulted in a true-positive finding, corresponding to six of eight (75%) false-positive findings and two of eight (25%) true-positive findings at per-vessel analysis and five of seven (71%) false-positive and two of seven (29%) true-positive findings at per-participant analysis. When nondiagnostic segments were excluded from analysis, CCTA correctly ruled out CAD in 37 participants, signifying that 37 of 68 (54%) participants would not have required additional ICA.

Table 3: Diagnostic Accuracy of UHR CCTA in Detection of Coronary Artery Disease in the Entire Study Population

In subgroup analysis, UHR CCTA showed the following results in the detection of coronary stenosis of at least 50%: Among participants with an Agatston score of at least 1000 (n = 24), sensitivity of 93% (13 of 14), specificity of 70% (seven of 10), and accuracy of 83% (20 of 24) were observed; however, for participants with prior stent implantation (n = 15), sensitivity of 100% (eight of eight), specificity of 86% (six of seven), and accuracy of 93% (14 of 15) were found. In participants with a known history of CAD, sensitivity of 100% (10 of 10), specificity of 88% (seven of eight), and accuracy of 94% (17 of 18) were observed. Detailed diagnostic test metrics in the detection of stenosis of at least 50% and in the detection of stenosis of at least 70% within these participant-based subgroups are provided in Table 4.

Table 4: Subgroup-based Diagnostic Performance of UHR CCTA in Detection of CAD

Assessment of Radiation Dose

The mean CT dose index volume and dose-length product, respectively, for the CT scans were as follows: calcium scoring CT (1.1 mGy ± 0.3, 22 mGy ∙ cm ± 6), UHR CCTA (67.7 mGy ± 19.2, 936 mGy ∙ cm ± 278), and aortoiliac CT angiography (4.6 mGy ± 1.2; 315 mGy ∙ cm ± 87), corresponding to mean effective doses of 0.3 mSv ± 0.1, 13.3 mSv ± 4.2, and 4.4 mSv ± 1.2, respectively.