Accuracy of dynamic stress CT myocardial perfusion in patients with suspected non-ST elevation myocardial infarction

This single-site prospective diagnostic-cross-sectional study was conducted at the Radboud University Medical Center, Nijmegen, the Netherlands. The CALAMARI (Ct Angiography/perfusion evaluation of non-st-eLevAtion MyocARdial Infarction) focusing on the sensitivity and specificity of CCTA with CT-MPI as the index test, compared to ICA with fractional flow reserve (FFR) in NSTEMI patients. The study protocol was compliant with the Declaration of Helsinki and received approval from the research ethics committee (NL71531.091.19). All participants provided written informed consent.

Adult patients (≥ 18 years old), who presented to the emergency cardiology department at our institution diagnosed with suspected NSTEMI regardless of Global Registry of Acute Coronary Events (GRACE) risk score and planned ICA within 72 h after admission were eligible for study. Study exclusion criteria were: (1) Hemodynamically or rhythmic unstable patients or patients with refractory angina pectoris, requiring emergent ICA according to treating cardiologist; (2) prior ST elevation myocardial infarction; (3) prior coronary artery bypass graft (CABG) surgery; (4) previous episode of contrast allergy or anaphylaxis after intravenous iodinated contrast administration; (5) contraindication for administration of both adenosine and regadenoson; (6) presence of pacemaker or ICD leads; (7) renal insufficiency defined as Glomerular Filtration Rate(GFR) < 45 ml/min/1.73m2; (8) pregnancy; (9) Body Mass Index > 35 kg/m². Myocardial infarction was defined as suspected type I myocardial infarction, according to the 2018 universal definition of myocardial infarction [12]. Based on an assumed prevalence of lesions of 65%, an estimated sensitivity of combined CTA and MPI of 89% and a marginal error of 8%, the study aimed to include 130 patients. Due to slow patient enrollment and competition with other studies, the study was terminated early.

Coronary CTA with CT-MPI was performed within 24 h before ICA. The study imaging protocol contained three examinations: Coronary Calcium score, CCTA and CT-MPI. All examinations were acquired using 320 row detection CT scanner (Aquilon One, Canon Medical Systems, Otawara, Japan). Patients were asked not to be taken caffeine-containing beverages immediately after written informed consent. The images and results of the three examinations were not used in clinical decision making and thereby did not affect in any way the care provided to the patient. Only incidental findings were reported to the responsible requesting physician.

Unenhanced cardiac CT for calcium scoring was obtained according to the scanning parameters in Table 1. Second CCTA was performed using prospective electrocardiogram triggered cardiac CT after administration of intravenous beta-blockers in case of heart frequency of > 60 beats/min up to a maximum of 10 mg metoprolol and sublingual nitroglycerine was given when no contra-indications were present. CCTA was obtained after administration 50–60 ml of 400 mg/ml iodinated contrast agent (Iomeron, Bracco Imaging, Cadempino, Switzerland) containing at an infusion rate of 4 to 4.5 ml/s, with bolus tracking in the descending aorta. All patients were scanned in one heartbeat during breath hold. The acquisition window was 70–80% R-R interval in most patients (n = 12), 35–55% R–R interval in 5 patients and 3 patient scans were performed with 70–99% R–R interval. All CT acquisition parameters are shown in Table 1. Hyperemia was induced by intravenous adenosine (standard dosage of 140 μg/kg/min) during a 4–5 min infusion. If the patient’s heart rate did not respond adequately in the first two minutes of infusion, this was increased to 210 μg/kg/min. If there was a contraindication for adenosine, regadenoson (intravenous bolus at fixed dose of 400 mcg/5 ml over 5 s) was used. The standard contrast injection protocol was a 50 ml of contrast bolus at 6 ml/s infusion rate (Iomeron, 400 mg l/ml), followed by 40 mL saline. The image acquisition was synchronized with the R-wave, occurring a few seconds before the contrast entered the left ventricle. The acquisition typically spanned 20–40 s, encompassing approximately 30 consecutive heartbeats. Continuous monitoring of cardiac rhythm and blood pressure was maintained. An experienced cardiovascular radiologist was at the bedside of the patient throughout the procedure.

All image reconstruction parameters are displayed in Table 1. Coronary calcium scores were calculated on 3 mm axial thickness reconstructions using the Agatston method [13] by an experienced cardiac radiographer at scanner's console. The CCTA images were reconstructed by using 3D reconstruction software (Terrarecon, Inc, Durham, NC, USA) to obtain central luminal line reformations (CPR) by one cardiovascular radiologist with more than 10 years of experience and one resident cardiology with 2 years of experience.

All CT-MPI images were reconstructed using the Advanced intelligent 9 Clear-IQ Engine (AiCE) [14] a deep learning-based algorithm (FC03/cardiac kernel, 8 mm slice thickness) [15] with the 4D similarity filter (SF) [16]. The 4D similarity filter provides noise reduction by averaging voxels corresponding to similar tissue types, resulting in more natural texture depiction with sharp vessel contours in comparison to the one obtained with conventional local spatial filtering [16].

The CT-MPI examinations were reconstructed and interpreted by two cardiovascular radiologists. One with 5–10 years of experience and one with 3–5 years of experience. Post-processing was performed using a dedicated workstation (Vitrea research 7.11.0, Vital Images, Minnetonka, MN, USA). The cardiovascular radiologist selected a target phase with optimal contrast enhancement in the left and right ventricles for segmentation. Cardiac axes and contours of the intra- and extravascular space were extracted automatically, with manual adjustments made if necessary, including axis selection, alignment, ventricular contouring, and identifying the highest value on the contrast inflow Time Density Curve before computing the results. For each dataset, five images in three imaging planes (four-chamber view, two-chamber view, and short axis views (basal, midventricular, and apical)) and one perfusion map were obtained.

Coronary CTA images including central luminal line reformations (CPR) were reviewed by two independent observers: One cardiovascular radiologist with more than 10 years of experience in radiology and one resident in cardiology with 2 years of experience in cardiology. All coronary segments with a diameter ≥ 1.5 mm were assessed according to a American Heart Association (AHA) 17-segment coronary artery model [17]. The severity of luminal diameter stenosis was evaluated according to the Coronary Artery Disease-Reporting and Data System (CAD-RADS) 2.0 classification, 0%: no visible stenosis, 1–24%: minimal stenosis, 25–49% mild, 50–69% moderate, 70–99% severe, and 100% occluded [18]. All coronary lesions were analyzed for the presence of high-risk plaque features (positive remodeling (PR), low attenuation plaque (LAP), spotty calcification (SC) and the napkin ring sign (NRS) defined as described in the CAD-RADS 2.0 classification [18]. A high-risk plaque was defined by the presence of at least 2 of these 4 plaque characteristics.

Image quality of CT-MPI was assessed by two cardiovascular radiologists. One with 5–10 years of experience and one with 5 years of experience. For each patient, a 16-segment perfusion map was evaluated at two different levels: per segment and per patient. Segment 17 was removed from the analysis due to poor segmentation. This was done on the same workstation for each reconstruction by using a single compartment model to calculate absolute myocardial blood flow (MBF) [19]. As a result, a total of 304 segments (19 cases × 16 segments) per acquisition method were assessed. A segment was considered normally perfused using a threshold of ≥ 1 ml/g/min. In case of < 1 ml/g/min in at least two adjacent segments [20], the vascular territory was considered positive for hypoperfusion. The patient's coronary anatomy obtained from CCTA was used by the software and controlled manually to assign myocardial perfusion defects to specific coronary vessels. The presence of significant CAD was determined per vessel territory based on the interpretation of available CCTA and CT-MPI images.

Radiation dose parameters were extracted from DICOM files of each investigation.

ICA was performed within 72 h after diagnosis and was conducted in accordance with local standards. Intermediate coronary lesions with diameter stenoses of at least 30% were assessed using intracoronary physiology (FFR and/or iFR) if it was considered technically feasible and safe by the interventional cardiologist. Evident culprit lesions (stenosis above 90% or angiographically suspicious for plaque rupture/thrombus) were not subject to FFR/iFR measurement but underwent percutaneous coronary intervention (PCI) with stenting if deemed appropriate. FFR measurements were performed with intracoronary adenosine. Significant CAD was defined as evident culprit lesion or an FFR value of ≤ 0.80 and/or iFR < 0.90. During ICA, images and FFR/iFR data were assessed immediately by the operator and subsequently reviewed by another independent interventional cardiologist (both more than 10 years of experience). The interventional cardiologists were blinded to the findings obtained from the CCTA and CT-MPI.

Demographic characteristics are presented as frequencies and percentages if it concerns categorical data, normally distributed continuous variables are reported as mean ± SD. The diagnostic performance of CCTA alone and CCTA with CT-MPI to identify significant obstructive CAD was evaluated on a per-vessel (primary analysis) and per-patient (secondary analysis) level, with invasive FFR as a reference standard. The diagnostic performance was reported as sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy with the 95% CI, with the ICA and FFR as reference. Diagnostic accuracy was defined as a proportion of accurate test results over the total test results. Statistical analyses were conducted using IBM SPSS statistics Version: 29.0.0.0. (Chicago, Illinois, USA) and R (R Foundation for Statistical Computing, Vienna, Austria).

Study population characteristics are shown in Table 2. Between January 2021 and September 2021, 20 patients (45% female) were enrolled and 19 completed all examinations and image quality for al 19 patients were deemed adequate for analysis. One patient was excluded from the analyses because of an allergic reaction on contrast medium during the CCTA. Throughout the study, no severe cardiac events or complications related to the study procedures were encountered during the CT-MPI or invasive FFR measurements.

The mean GRACE risk score was 110 ± 20.3 meaning intermediate risk. Two patients (10%) had previously undergone PCI for stable CAD. Nine patients had single vessel CAD, while 7 had multi-vessel CAD (Table 2). Significant CAD was observed in 16 patients (80%), with a culprit lesion (n = 14) or a FFR of less than 0.8 (n = 2). Among these 16 patients, 15 underwent immediate PCI, and one patient required an emergent CABG procedure. In contrast, four patients did not exhibit obstructive CAD despite increase in troponin levels. A total of 17 lesions underwent FFR assessment, of which 13 were FFR- negative. The remaining 4 FFR positive lesions were treated with either PCI (n = 3) or CABG (n = 1). In 2 patients, the infarct-related artery was not immediately identifiable, and positive FFR measurements were used to guide revascularization decisions. A case example including all examinations during the study is shown in Fig. 1.

The mean dose length product (DLP (mGy.cm)) are shown in Table 1. The mean DLP for calcium score, CCTA and CT-MPI are 61.2 (± 11.4), 132.6 (± 81.2) and 186.7 (± 70.4) mGy.cm respectively. The mean total radiation dose per patient is 380.5 (± 140.4) mGy.cm.

The mean calcium score was 771 meaning severe atherosclerosis. There was no significant difference in the interpretation of the CT-MPI results between the two cardiovascular radiologists. All CCTAs were interpretable (CADRADS N, n = 0). A total of 98 lesions were identified and analyzed, with most of them being calcified plaques (63%). Among the analyzed lesions, 51 exhibited stenoses ≥ 50% (CADRADS 3 or higher), while 47 had stenoses below 50% (CADRADS < 3). A comparison of anatomical severity based on CCTA and ICA is shown in Fig. 2. CCTA showed more often a multivessel disease than ICA + FFR, with two vessel disease (n = 8 vs n = 5) and three vessel disease (n = 3 vs n = 2) respectively. The modifier high risk plaque (2 or more HRP features) was observed in 6 (30%) patients. The most common were positive remodeling (n = 17) and low attenuation plaque (n = 8).

All patients had a hemodynamically adequate stress response on adenosine. CT-MPI shows good correlation with the results of ICA in 17 out of the 19 patients and in 55 out of the 57 lesions (Table 3). Coronary CTA with ≥ 50% stenosis demonstrated a per vessel sensitivity and specificity for the detection of significant stenosis of 100% (95% CI: 86–100%) and 75% (95% CI: 58–88%) respectively. Coronary CTA combined with CT-MPI showed a sensitivity of 90% (95% CI: 70–99%) but a higher specificity of 100% (95% CI: 90%−100%) (Table 4). By adding CT-MPI to CCTA, the positive predicted value (PPV) increases to 100% (95% CI: 82–100) versus 73% (95% CI: 58–88) for CCTA alone, without change in the negative predicted value 95% (95% CI: 82–99) versus 100% (95% CI: 87–100), respectively. This results in more patients being correctly classified when CCTA and CT-MPI are combined, namely 96% (95% CI: 88–100), versus CCTA alone 84% (95% CI: 73–93).