Abstract
Background
Total atherosclerotic plaque burden assessment by CT angiography (CTA) is a promising tool for diagnosis and prognosis of coronary artery disease (CAD) but its validation is restricted to small clinical studies. We tested the feasibility of semi-automatically derived coronary atheroma burden assessment for identifying patients with hemodynamically significant CAD in a large cohort of patients with heterogenous characteristics.
Methods
This study focused on the CTA component of the CORE320 study population. A semi-automated contour detection algorithm quantified total coronary atheroma volume defined as the difference between vessel and lumen volume. Percent atheroma volume (PAV = [total atheroma volume/total vessel volume]×100) was the primary metric for assessment (n=374). The area under the receiver operating characteristic curve (AUC) determined the diagnostic accuracy for identifying patients with hemodynamically significant CAD defined as ≥50% stenosis by quantitative coronary angiography and associated myocardial perfusion abnormality by SPECT.
Results
Of 374 patients, 139 (37%) had hemodynamically significant CAD. The AUC for PAV was 0.78 (95% confidence interval [CI] 0.73–0.83) compared to 0.84 [0.79–0.88] by standard expert CTA interpretation (p=0.02). Accuracy for both CTA (0.91 [0.87, 0.96]) and PAV (0.86 [0.81–0.91]) increased after excluding patients with history of CAD (p<0.01 for both). Bland-Altman analysis revealed good agreement between two observers ( bias of 280.2 mm3 [161.8, 398.7]).
Conclusions
A semi-automatically derived index of total coronary atheroma volume yields good accuracy for identifying patients with hemodynamically significant CAD, though marginally inferior to CTA expert reading. These results convey promise for rapid, reliable evaluation of clinically relevant CAD.
INTRODUCTION
Semi-automated contour detection algorithms based on CT coronary angiography (CTA) have shown promise for assessment of total coronary atherosclerotic plaque burden which may provide fast, reliable assessment of coronary artery disease (CAD).1,2 However, results are available only from relatively small clinical studies which focused on the main coronary arteries sparing side branches. In a recent study, evaluation of total atheroma burden by CTA performed better than standard lumen assessment for identifying coronary arteries with hemodynamically significant coronary arterial stenosis as determined by fractional flow reserve (FFR).3 These results emphasize the importance of atherosclerotic burden for determining coronary blood flow restriction. Indeed, lower grade stenoses may cause critical reduction in coronary blood flow in the presence of diffuse atherosclerotic disease.4 On the other hand, FFR results may not correspond to myocardial perfusion data, which are more commonly used in clinical practice to guide medical management.5 The purpose of this study was to test the feasibility and the diagnostic accuracy of total coronary atheroma assessment obtained by a semi-automated contour detection algorithm for identifying coronary artery disease associated with myocardial perfusion abnormalities in a large clinical cohort.
METHODS
Study design and study population
The study design of the CORE320 multicenter study has been previously detailed.6 The CORE320 study—Coronary Artery Evaluation using 320-row Multidetector Computed Tomography Angiography and Myocardial Perfusion—is a prospective, multicenter, multinational, diagnostic study designed to compare the accuracy of combined CTA and myocardial computed tomography perfusion imaging (CTP) against the combination of invasive coronary angiography (ICA) and Single-Photon Emission Computed Tomography Myocardial Perfusion Imaging (SPECT-MPI).7 Patients 45 to 85 years of age who were referred for clinically indicated ICA for suspected or known CAD were enrolled.
CT acquisition, image reconstruction, transfer, and analysis
A detailed description of CORE320-image acquisition and interpretation methods has been published.8 In brief, all CT images were acquired before invasive angiography using a single protocol developed for a 320 × 0.5 mm-detector row CT system (Aquilion ONE, Toshiba Medical Systems, Otawara, Japan). Patient preparation included oral (75–150 mg) or IV (up to 15 mg) metoprolol and sublingual, fast-acting nitrates. Fifty to 70 mL of iodinated contrast (Iopamidol 370 mg iodine/mL) was injected intravenously at 4.0–5.0 mL/sec for each of the separate, axial, prospectively ECG-triggered acquisitions. For all CTA acquisitions, de-identified sinograms were reconstructed, processed, and interpreted by independent core laboratories. CT data were reconstructed to generate 0.5-mm slice thickness images with a 0.25-mm increment using both a standard (FC43) and a sharp (FC05) convolution kernel. Two level III certified investigators evaluated each CTA study for the presence and severity of CAD; disagreements were resolved by consensus. Readers examined all coronary artery segments of 1.0 mm in diameter or more for the presence of CAD using a 19-segment coronary artery model. All coronary lesions with a subjective diameter stenosis of ≥30% underwent quantitative evaluation on a continuous scale (0–100%) using software tools (Vitrea™ FX version 3.0 workstation, Vital Images, Minnetonka, MN, USA) at the discretion of the reader. Typical analysis time for each study was 35–45 minutes (range 15–90 minutes). The analysis was shorter in patients without known CAD (on average, 25–35 minutes).
Coronary atheroma volume analysis
All reconstructed datasets were transferred to an offline workstation for quantitative coronary atheroma volume analysis using dedicated software with a semi-automated 3-dimensional (3D) contour detection algorithm (QAngio CT Research Edition version 2.0 RC4, Medical imaging systems (MEDIS), Leiden, the Netherlands).2,9,10 The software was selected for our purposes based on its validation record and its user friendly interface. Quantitative atheroma analysis was performed by two independent, experienced observers who were blinded to initial CTA, quantitative coronary angiography (QCA), and clinical data. The reconstructed images were displayed at a window width of 700 and a window level of 200 to standardize quantitative coronary artery assessment. On the basis of longitudinal contours, cross-sectional images at 0.5-mm intervals were obtained to create transversal lumen and vessel wall contours, using automated contour detection techniques applied to the intensity gradients in the cross sections and guided by the longitudinal contours. These cross-sectional contours were examined and, if necessary, corrected by the observer (Figure 1). All coronary vessels were assessed using a 19-coronary segment model, including each epicardial vessel and side branches with at least 1.5 mm in diameter. Therefore, we included smaller segments than in previous studies given the advancements in CT technology in recent years. Segments containing stents and those with poor image quality were excluded from analysis. The following parameters were derived per segment: vessel length, vessel volume, lumen volume, mean plaque burden, minimum and maximum plaque thickness, and mean lumen and plaque intensity. The plaque volume was calculated by subtracting the lumen volume from the vessel volume. For each patient, the vessel, lumen, atheroma, and length values were calculated by adding all the analyzed segments. Required time for analysis (including editing) was typically 20–30 minutes per patient (range 10–60 minutes). Mean analysis time was shorter in patients without history of CAD, typically 15–25 minutes.
Figure 1. Plaque Assessment by CT Angiography Using a Semi-Automated Contour Detection Algorithm (Panels A–D).
(A) A CT image of the left anterior descending artery is shown using a curved multi-planar reformatted (cMPR) volume.
(B) The same vessel as in A is shown after lumen and vessel wall contour tracking by the software. The dotted line indicates the site of cross-sectional image display in panels C and D.
(C) A Cross-sectional image at the in B specified site is shown. Noncalcified atherosclerotic plaque is noted resulting in moderate lumen encroachment.
(D) The same image as in C is shown after contour tracing of the vessel and lumen borders which define the plaque area (vessel area minus lumen area).
Based on prior investigations of plaque assessment, we defined three quantitative atheroma volume indices11:
Percent atheroma volume (PAV), %: (total atheroma volume/total vessel volume)×100.
Length normalized total atheroma volume (length normTAV), mm3/m: total atheroma volume/total segment length.
Normalized total atheroma volume (normTAV), mm3: (total atheroma volume/total segment length)×mean total vessel length; mean total vessel length = average length of total vessel in the sample.
Total lumen volume was also indexed to total segment length and LV mass.
Invasive coronary angiography acquisition and analysis
Invasive angiography was performed using standard techniques within 60 days following SPECT and CTA acquisition. Quantitative coronary angiography (QCA) was performed using standard, validated analysis software (CAAS II QCA Research version 2.0.1, PIE Medical Imaging, Maastricht, The Netherlands). Coronary segments were defined using a 19-coronary segment model, and all coronary segments 1.5 mm or more in diameter were analyzed quantitatively. Significant coronary artery stenosis (obstructive CAD) was defined as ≥50% diameter stenosis by QCA.
SPECT acquisition and analysis
Myocardial territories were analyzed by SPECT for rest and stress myocardial perfusion abnormalities with a severity and reversibility-scored, 4-point scoring model using a 13-territory model.12 The summed stress score (SSS) was defined as the sum of abnormal myocardial segments at stress phase.7,8 In the analysis, artifacts did not contribute to the summed stress score (SSS) and therefore a SSS ≥1 defined an abnormal SPECT study according to established methods of multi-center core SPECT laboratories.13
Definition of outcome
Study outcome was the presence of at least one ≥50% stenosis by QCA with an associated perfusion abnormality SPECT (SSS ≥1).7,12 Ambiguities in regards to associations were resolved by a dedicated consensus committee compromised of the respective core lab leaders.
Statistical analysis
Patients’ data were summarized by using median and interquartile range for continuous variables and percent for categorical variables. Receiver operating characteristic (ROC) curves were used to determine the diagnostic performance of CTA stenosis and total coronary atheroma volume indices to predict flow-limiting coronary stenosis as defined by the combination of QCA and SPECT. We compared the area under the ROC curve (AUC) for total coronary atheroma volume indices with standard expert CTA stenosis assessment for identifying patients with significant CAD and associated perfusion abnormality, using standard significance tests for paired data. In addition, 95% confidence intervalles were computed for differences between AUC values using the bootstrap with 2000 samples. AUC values were compared using standard methods. Pre-specified sub-analyses included patients without history of coronary artery disease and restriction of plaque volume analysis to “core” coronary segments, i.e., proximal coronary segments of the three coronary arteries (proximal and mid right coronary artery, left main, proximal and mid left anterior descending coronary artery, and proximal left circumflex coronary artery. All data were reported with 95% confidence intervals (CI). The 381 CT studies were analyzed for coronary atheroma indices by two observers who read 198 and 183 studies, respectively. The inter-observer agreement for total coronary atheroma volume was determined for 50 randomly selected patients which were read by both observers; Bland-Altman plots were conducted to assess bias and inter-observer variability. Statistical significance was determined at p values <0.05. All analyses were conducted using the statistical software packages SAS (version 9.3 for Windows, SAS Institute Inc., Cary, NC, USA) and Stata (version 12 for Windows, StataCorp, College Station, TX, USA).
RESULTS
Clinical characteristics
Of 381 patients included in the final CORE320 study sample, seven patients were excluded from this analysis because of poor overall image quality, resulting in a final study population of 374 patients. A total of 6,032 coronary arterial segments were evaluated. Of these, 189 (3.1%) were excluded for analysis because they were uninterpretable by CT because of motion and/or poor contrast to noise ratio. Coronary calcium was not a reason for exclusion. Table 1 displays the clinical data and baseline characteristics for the entire study population and according to the presence of at least one ≥50% stenosis by QCA and associated myocardial perfusion abnormality. The median age of all patients was 62 years (range 56 to 68 years): 67% were male, 33% were Asian, 11% were African-American, and 56% were Caucasian. Patients had a high prevalence of risk factors (hypertension 78%, diabetes 34%, dyslipidemia 68%, current smoker 18%, and previous percutaneous coronary intervention 29%). The proportion of stable angina was 97% vs. 3% unstable angina. Median calcium score was 161 Interquartile Range [IQR 9, 509] for all patients, 251 (68, 579) in patients with known CAD and 111 [1, 404] in patients without prior history of CAD.
Table 1.
Baseline Patients Characteristics
| Characteristic | All n = 374 | Stenosis† n = 139 | No stenosis n = 235 | p-value |
|---|---|---|---|---|
| Age, in years | 62.1 (55.7–68.4) | 62.0 (55.6–68.4) | 62.1 (55.7–68.5) | 0.62 |
| Male Gender, n (%) | 249 (67) | 112 (81) | 137 (58) | <0.0001 |
| Asian, n (%) | 122 (33) | 47 (34) | 75 (32) | 0.78 |
| African American, n (%) | 43 (11) | 14 (10) | 29 (12) | |
| Caucasian, n (%) | 209 (56) | 78 (56) | 131 (56) | |
| Weight, Kg | 74 (65–86) | 72 (64–86) | 74 (65–85) | 0.61 |
| Height, cm | 167 (160–173) | 168 (161–174) | 166 (158–172) | 0.06 |
| BMI, kg/m2 | 26.6 (24.1–30.1) | 26.2 (24.0–28.7) | 26.7 (24.2–30.8) | 0.08 |
| Hypertension, n (%) | 290 (78) | 114 (83) | 176 (75) | 0.10 |
| Diabetes, n (%) | 126 (34) | 55 (40) | 71 (30) | 0.06 |
| Dyslipidemia, n (%) | 248 (68) | 102 (76) | 146 (63) | 0.01 |
| Previous MI, n (%) | 99 (26) | 61 (44) | 38 (16) | <0.0001 |
| Prior percutaneous coronary intervention, n (%) | 111 (30) | 59 (42) | 52 (22) | <0.0001 |
| Currently Smokes, n (%) | 63 (18) | 19 (15) | 44 (19) | 0.15 |
| Past Smoker, n (%) | 131 (37) | 56 (43) | 75 (33) | |
| Never Smokes, n (%) | 163 (46) | 55 (42) | 108 (48) | |
| Family history of CAD, n (%) | 157 (44) | 61 (47) | 96 (43) | 0.54 |
| Previous heart failure, n (%) | ||||
| NYHA class I | 8 (16) | 3 (14) | 5 (18) | 0.63 |
| NYHA class II | 40 (82) | 18 (86) | 22 (79) | |
| NYHA class III | 1 (2) | 0 (0) | 1 (4) | |
| NYHA class IV | 0 (0) | 0 (0) | 0 (0) | |
| Angina at presentation | ||||
| Unstable angina, n (%) | 9 (3) | 3 (3) | 6 (4) | 0.63 |
| Stable angina, n (%) | 251 (97) | 104 (97) | 147 (96) | |
| Angina (within 30 days), Canadian Class, n (%) | ||||
| 0 | 61 (22) | 15 (13) | 46 (27) | 0.005 |
| 1 | 110 (39) | 39 (35) | 71 (42) | |
| 2 | 97 (34) | 51 (45) | 46 (27) | |
| 3 | 12 (4) | 6 (5) | 6 (4) | |
| 4 | 3 (1) | 2 (2) | 1 (1) | |
| Prior Stress Testing (30 days), n (%) | ||||
| ECG only | 15 (4) | 7 (5) | 8 (3) | 0.44 |
| Echo | 8 (2) | 5 (4) | 3 (1) | 0.13 |
| Prior Stress Testing Result, n (%) | ||||
| Positive | 11 (50) | 5 (42) | 6 (60) | 0.39 |
| Negative/equivocal | 11 (50) | 7 (58) | 4 (40) | |
| LV mass*, g | 148 (128–175) | 166 (134–186) | 144 (126–162) | 0.0001 |
| LV mass index*, g/m2.7 | 36.7 (32.6–43.2) | 38.7 (33.6–46.7) | 36.1 (32.0–41.4) | 0.003 |
| Coronary artery stenosis (CTA≥50%), n (%) | 246 (66) | 129 (93) | 117 (50) | <0.0001 |
| Coronary artery stenosis (QCA≥50%), n (%) | 223 (60) | 132 (95) | 91 (39) | <0.0001 |
| Myocardial hypo-perfusion (SPECT: SSS>0), n (%) | 187 (50) | 139 (100) | 48 (20) | <0.0001 |
| Myocardial ischemia (SPECT: SDS>0), n (%) | 148 (40) | 109 (78) | 39 (17) | <0.0001 |
Continuous variable data are presented as median (interquartile range).
Incomplete data for LV mass: n=332 for all, n=122 for stenosis, and n=210 for no stenosis.
Stenosis refers to at least one ≥50% stenosis by QCA with related perfusion deficit by SPECT.
BMI = body mass index; CAD = coronary artery disease; CTA = computed tomography angiography; LV = left ventricular; MI= myocardial infarction; QCA = quantitative coronary angiography; SPECT = single-photon emission computed tomography; SSS = summed stress score; SDS = summed difference score.
Observer Variability for Plaque Volume Assessment by CTA
Bland-Altman analysis in 49 patients (one study result out of 50 was not available for technical reasons) revealed good agreement as shown in Figure 2. A bias was found of 280.2 mm3 (95% CI: 161.8, 398.7) with a ratio of standard deviations of 1.17 (95% CI: 1.01, 1.35). Based on the bias found, the final results (vessel and lumen volume measurements) were adjusted using linear calibration curves with the more senior observer as the reference standard.
Figure 2. Variability of Atheroma Volume Assessment among Observers.
Shown are Bland-Altman plots on the difference of total atheroma volume between observer 1 and observer 2 in 49 patients. Bias was 280.20 mm3 (95% CI: 161.8, 398.7 with a ratio of standard deviations of 1.17 (95% CI: 1.01, 1.35)).
Coronary artery disease evaluation
The measures of atheroma burden in subjects stratified according to the reference standard incidence of coronary stenoses with myocardial perfusion abnormalities are shown in Table 2. Of 374 patients, 139 patients (37.2%) had flow-limiting CAD. The median CTA stenosis, length normTAV, PAV, and normTAV were 62%, 6.0 m3/m, 55.0%, and 2962.0 mm3, respectively (Table 2).
Table 2.
Quantified Coronary Artery Characteristics
| Characteristic | All n = 374 | Stenosis‡ n = 139 | No Stenosis n = 235 | p-value |
|---|---|---|---|---|
| Vessel length, mm | 505.0 (420.0–570.6) | 463.5 (385.3–552.9) | 525.0 (451.5–582.4) | <0.0001 |
| Vessel volume, mm3 | 5388.4 (4187.1–6665.0) | 5282.9 (4034.9–6493.2) | 5458.9 (4327.6–6742.0) | 0.22 |
| Lumen volume, mm3 | 2373.4 (1799.8–3002.0) | 2083.0 (1583.0–2752.7) | 2574.1 (1964.4–3154.7) | <0.0001 |
| Plaque volume, mm3 | 2946.4 (2317.8–3615.0) | 3099.5 (2407.8–3745.4) | 2864.4 (2223.8–3535.2) | 0.02 |
| Percent atheroma volume, % | 55.0 (50.4–60.0) | 59.2 (55.6–63.5) | 52.4 (48.3–57.5) | <0.0001 |
| Normalized TAV, mm3 | 2962.0 (2520.4–3409.6) | 3321.6 (2902.7–3765.5) | 2804.2 (2355.4–3186.4) | <0.0001 |
| Length normalized TAV, mm3/m | 6.0 (5.1–6.9) | 6.7 (5.9–7.6) | 5.7 (4.8–6.5) | <0.0001 |
| Core Vessel length*, mm | 139.1 (115.5–162.0) | 129.5 (103.0–150.0) | 148.0 (124.5–167.5) | <0.0001 |
| Core Vessel volume*, mm3 | 2388.9 (1906.4–3049.4) | 2212.8 (1808.2–2927.4) | 2480.3 (1950.1–3101.2) | 0.02 |
| Core Lumen volume*, mm3 | 1025.9 (785.0–1432.0) | 856.6 (714.4–1172.8) | 1145.3 (889.0–1497.1) | <0.0001 |
| Core Plaque volume*, mm3 | 1341.9 (1030.6–1611.0) | 1343.4 (1067.4–1654.3) | 1332.1 (1022.1–1590.8) | 0.35 |
| Core Percent atheroma volume†, % | 54.5 (49.9–60.4) | 59.4 (53.9–64.8) | 52.5 (47.6–57.2) | <0.0001 |
| Core Normalized TAV*, mm3 | 1356.9 (1141.9–1570.3) | 1493.7 (1296.4–1756.9) | 1266.8 (1094.9–1465.2) | <0.0001 |
| Core Length normalized TAV*, mm3/m | 9.8 (8.3–11.4) | 10.8 (9.4–12.7) | 9.2 (7.9–10.6) | <0.0001 |
| Lumen volume/length, mm3/m | 4.8 (4.1–5.7) | 4.5 (3.8–5.1) | 5.0 (4.2–5.8) | 0.0004 |
| Lumen volume/length/LV mass†, mm3/m/g | 0.032 (0.026–0.041) | 0.028 (0.023–0.035) | 0.035 (0.028–0.042) | <0.0001 |
| Coronary artery stenosis by CTA, % | 62 (41–89) | 91 (69–100) | 48 (36–67) | <0.0001 |
| Coronary artery stenosis by QCA, % | 60 (28–91) | 95 (76–100) | 43 (15–63) | <0.0001 |
| Calcium Score, Agatston score | 161 (9–509) | 373 (143–936) | 64 (0–300) | <0.0001 |
Continuous variable data are presented as median (interquartile range).
Core segments are proximal and mid right coronary artery, left main, proximal and left anterior descending coronary artery, and proximal left circumflex coronary artery.
Incomplete data for LV mass: n=332 for all, n=67 for revascularization, and n=265 for no revascularization.
Stenosis refers to at least one ≥50% stenosis by QCA with related perfusion deficit by SPECT;
CTA = computed tomography angiography; LV = left ventricular; TAV = total atheroma volume; QCA = quantitative coronary angiography.
Diagnostic performance of CTA stenosis and quantitative total coronary atheroma volume indices for pre-detection of flow-limiting coronary stenosis
Table 3 and Figure 3 show the diagnostic accuracies for CTA and atheroma indices. The AUC for predicting flow-limiting coronary stenosis for standard CTA stenosis assessment was 0.84 (95% CI, 0.79–0.88) compared to 0.78 (0.73–0.83) for PAV (p=0.02). The difference for AUC between CTA and PAV was 0.057 (95% bootstrap CI: 0.008,0.104). After excluding patients with stents AUC rose to 0.83 for PAV (0.78–0.88). The AUC for normTAV (0.72 [0.67–0.78]) was lower than that by PAV and CTA (p<0.001). When only the proximal coronary segments were considered in a subanalyis, the results were similar to that reported above for the entire coronary tree (Figure 3B). Accuracies were also similar when a perfusion defect by SPECT was not informed by the location of a stenoses by QCA, i.e., and endpoint of any 50% QCA stenosis in the presence of any perfusion defect by SPECT in the same patient regardless of their location: AUC 0.85 (0.81–0.89) for CTA and 0.79 for PAV ((0.75–0.84), p=0.03). The difference for AUC between CTA and PAV was 0.051 (95% bootstrap CI: −0.005,0.105). Information by PAV or normTAV did not significantly improve AUC for CTA (Figures 3C and 3D). When separating the combined endpoint, AUC was greater for PAV vs. QCA alone ((0.85 [0.81–0.88]) than PAV vs. SPECT alone (0.67 [0.62–0.67], p<0.001). Of note, in 85 patients SPECT findings were normal despite ≥50% stenosis by QCA while only 44 patients with abnormal SPECT findings had <50% stenosis by QCA.
Table 3.
Diagnostic Accuracy According to History of Coronary Artery Disease
| Overall | No history of CAD | History of CAD | |||||
|---|---|---|---|---|---|---|---|
|
| |||||||
| Effect | AUC | 95% CI | AUC | 95% CI | AUC | 95% CI | p |
| Percent atheroma volume, % | 0.78 | 0.73, 0.83 | 0.86 | 0.81, 0.91 | 0.60 | 0.51, 0.69 | <.001 |
| Normalized TAV, mm3 | 0.72 | 0.67, 0.78 | 0.80 | 0.74, 0.86 | 0.60 | 0.50, 0.69 | <.001 |
| Length normalized TAV, mm3/m | 0.72 | 0.67, 0.78 | 0.80 | 0.74, 0.86 | 0.60 | 0.50, 0.69 | <.001 |
| CTA stenosis, % | 0.84 | 0.79, 0.88 | 0.91 | 0.87, 0.96 | 0.68 | 0.59, 0.77 | <.001 |
| Core percent atheroma volume*, % | 0.75 | 0.69, 0.80 | 0.83 | 0.77, 0.89 | 0.60 | 0.51, 0.69 | <.001 |
| Core normalized TAV*, mm3 | 0.71 | 0.65, 0.76 | 0.75 | 0.68, 0.82 | 0.63 | 0.54, 0.72 | 0.036 |
| Core length normalized TAV*, mm3/m | 0.71 | 0.65, 0.76 | 0.75 | 0.68, 0.82 | 0.63 | 0.54, 0.72 | 0.036 |
| Lumen volume/length, mm3/m | 0.61 | 0.55, 0.67 | 0.63 | 0.54, 0.71 | 0.54 | 0.44, 0.63 | 0.174 |
| Lumen volume/length/LV mass†, mm3/m/g | 0.67 | 0.60, 0.73 | 0.68 | 0.59, 0.77 | 0.59 | 0.49, 0.69 | 0.181 |
| Core Lumen volume*, mm3 | 0.67 | 0.61, 0.73 | 0.66 | 0.58, 0.75 | 0.59 | 0.49, 0.69 | 0.276 |
Core segments definition is the same as in Table 2.
Incomplete data for LV mass: n=332 for all, n=67 for revascularization, and n=265 for no revascularization.
AUC = area under the receiver operating characteristic curve; CAD = coronary artery disease; CI = confidence interval; CTA = computed tomography angiography; LV = left ventricular; TAV = total atheroma volume.
Figure 3. Diagnostic Performance of Total Coronary Atheroma Volume Evaluation Compared with Standard CTA Assessment.
Panel A shows the ROC curves for percent atheroma volume (AUC, 0.78; 95% CI, 0.73–0.83), normalized total atheroma volume (AUC, 0.73; 95% CI, 0.67–0.78), and CTA stenosis (AUC, 0.84; 95% CI, 0.79–0.88) for identifying a patient with ≥50% stenosis by quantitative coronary angiography and associated myocardial perfusion abnormality by SPECT. There was a significant difference in AUC between the percent atheroma volume curve and the CTA stenosis curve (p=0.02) and normalized total atheroma volume (p=0.04). There was also a significant difference in AUC between normalized total atheroma volume curve and CTA stenosis curve (p=0.001).
Panel B shows the same analysis as in Panel A but restricted to proximal coronary arterial segments (left main coronary artery, proximal and mid left anterior descending artery, proximal and mid right coronary artery, and proximal left circumflex coronary artery). There was a significant difference in AUC between core normalized total atheroma volume curve and CTA stenosis curve (p<0.001) and between core CTA stenosis and core percent atheroma volume curve (p=.0007), however a significant difference was not found between core percent atheroma volume and core normalized total atheroma volume (p=0.23 ).
Panel C shows the diagnostic accuracy when information from CTA is combined with percent atheroma volume or normalized total atheroma volume, respectively. There is no statistically significant increase in diagnostic accuracy for either index over standard CTA evaluation.
Panel D shows the same analysis as in Panel C but restricted to proximal coronary segments. There is no statistically significant increase in diagnostic accuracy for either index over standard CTA evaluation.
Abbreviations: AUC = area under the receiver operating characteristic curve; CI = confidence interval; CTA = computed tomography angiography; ROC = receiver operating characteristic; SPECT: single photon emission tomography; PAV: percent atheroma volume; NTAV: normalized total atheroma volume.
Accuracy for patients without previous coronary artery disease
In patients without prior history of CAD, the diagnostic accuracy among the tested atheroma indices as well as the standard CTA assessment by expert readers was higher than for the entire cohort (Table 3). The AUC for PAV (0.86; CI, 0.81–0.91) was non-significantly lower than that from the expert-determined CTA stenosis evaluation (0.91; 95% CI, 0.87–0.96, p=0.08). AUCs of normTAV and length normTAV (0.80; 95% CI, 0.74–0.86) were smaller than for expert-determined CTA stenosis (Table 3). In patients with history of CAD, the AUCs of all parameters were less than 0.70, indicating modest performance.
DISCUSSION
A semi-automated derived index of percent atheroma volume (PAV) obtained by CTA identified patients with hemodynamically significant CAD with good diagnostic accuracy. Results were slightly inferior, however, compared to expert readers who assessed the CTA data sets by conventional methods. Reader experience has been shown to influence the results of standard CTA assessment.14 High inter-observer agreement for the semi-autoimated PAV evaluation raises hope for achieving consistent results among readers with a broad range of experiences even though the small reported bias among observers suggests some degree of experience may be indicated.15 The software required little contour editing to achieve the reported performance, which resulted in shorter analysis times compared to standard CTA stenosis assessment. Our analysis was not restricted to proximal coronary segments but included the entire coronary tree. Thus, our results reveal promising diagnostic capabilities for plaque volume assessment by CTA which can be achieved in a rapid, reliable fashion.
While diagnostic accuracies of PAV - as well those by standard CTA - were solid for the entire study population, performance increased to good levels when only patients without prior CAD were considered. Since patients without prior CAD are the current target population for CTA assessment, clinical application of PAV may facilitate practical, reliable assessment for hemodynamically significant CAD similar to what has been achieved with noninvasive fractional flow reserve assessment by CT.16 Furthermore, given the evidence for total plaque volume as an important risk factor for adverse coronary events,17 it is conceivable that the prognostic information of such assessment will further enhance the clinical value of PAV evaluation. Several recent studies addressed the diagnostic value of automated contour detection algorithms by CTA. In comparison to intravascular ultrasound (IVUS), quantitative coronary atheroma assessment by CTA yielded good agreement1,18 Using percent aggregate plaque volume – which is comparable to PAV in the current study - a recent clinical study found superior diagnostic accuracy compared to standard lumen assessment by CTA for identifying hemodynamically significant coronary lesions.3 Our results principally confirm the promising performance of plaque volume assessment by CTA, but extend its significance by assessing the entire coronary tree instead of studying individual lesions and by using myocardial perfusion imaging instead of FFR for hemodynamic assessment. While the reference standard for optimal CAD assessment remains controversial,19 comprehensive myocardial blood flow assessment in addition to anatomic information carries some advantages over lesion specific evaluation.5 Because atherosclerotic plaque burden influences coronary blood flow in regard to its quantity and distribution, it may be better positioned to address the hemodynamic impact of CAD.20 Future investigations will define the role of coronary atheroma quantification for clinical management, particularly, in comparison to the many emerging tools available with cardiac CT, e.g, myocardial perfusion imaging, CT-FFR, or transluminal attenuation gradient assessment.21
Limitations
Our study has limitations. First, the CORE320 study was not designed to address the present research question and thus a power calculation for this analysis was not performed. As power calculations would not be appropriate as a post hoc measure, we provide 95% confidence intervals for our analyses supporting the robustness of our findings. It should also be noted that the CORE320 study population contains patients who are at higher risk than typically seen with the application of CTA. Thus, results may not be applicable to low risk populations. Second, prospectively total atheroma assessment was limited to non-stented segments. Third, we excluded coronary artery segments of less than 1.5 mm diameter for plaque assessment because of CT’s modest spatial resolution in comparison with ICA. Fourth, we recognize that this study does not assess the incremental value of plaque volume analyses when combined with CTA, or with CTA plus CT myocardial perfusion imaging. However, the purpose of this study was to establish the diagnostic accuracy of a rapid, practical tool for CAD assessment – without the need for additional imaging. Fifth, the contour detection algorithm undergoes frequent upgrades and improvements, some of them not included in our version. For example, we noted that the software computed substantial “plaque” volumes even in patients without visually apparent CAD, indicating that these data had high residual noise levels. Further improvements in software performance may allow trimming such noise for the benefit of better diagnostic performance. Similarly, there is hope that further software upgrades may allow including smaller segments, stented lesions, and perform better with poor image quality or extensive calcification.
Lastly, it should be noted that the software is for research purposes only at this time and not yet validated for clinical use.
Conclusions
In patients without history of coronary artery disease, a semi-automatically derived index of percent atheroma volume by CTA yields good accuracy for identifying obstructive coronary artery disease that is is associated with myocardial perfusion abnormalities. These results, while slightly inferior to expert evaluation, convey promise for a rapid, reliable assessment of clinically relevant coronary artery disease. Future software developments are likely to expand the performance and the role of automated assessments of CAD by CTA for clinical practice.
Highlights.
Percent coronary atheroma volume can be obtained by CT angiography with similar processing times as stenosis assessment by expert readers
Percent coronary atheroma volume by CT angiography identifies patients with obstructive coronary artery disease and associated myocardial perfusion abnormalities with good accuracy
Percent coronary atheroma volume is more accurate than other atherosclerotic indices for identifying patients with obstructive coronary artery disease and associated myocardial perfusion abnormalities
Accuracy of semi-automatedly derived percent coronary atheroma volume by CT angiography for identifying patients with obstructive coronary artery disease and associated myocardial perfusion abnormalities is similarly high as with CT stenosis assessment by expert readers in patients without history of coronary artery disease
Acknowledgments
Funding Sources
The sponsor of the CORE320 study, Toshiba Medical Systems Corporation, was not involved during any stage of the planning, design, data acquisition, data analysis, or manuscript preparation of this study.
Footnotes
Disclosures
Dr. J.H.C. Reiber has a part-time appointment as Prof of Medical Imaging at the Leiden University Medical Center (LUMC) and is the CEO of Medis medical imaging systems, Leiden, the Netherlands.
Dr. P. Kitslaar has a research appointment at the LUMC, Division of Image Processing (LKEB), Dept of Radiology and is an employee of Medis, Leiden, the Netherlands.
Drs Rybicki and Arbab-Zadeh disclose their membership of the CORE320 steering committee. The CORE320 study was sponsored by Toshiba Medical Systems.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References
- 1.Boogers MJ, Broersen A, van Velzen JE, et al. Automated quantification of coronary plaque with computed tomography: Comparison with intravascular ultrasound using a dedicated registration algorithm for fusion-based quantification. Eur Heart J. 2012;33(8):1007–1016. doi: 10.1093/eurheartj/ehr465. [DOI] [PubMed] [Google Scholar]
- 2.Boogers MJ, Schuijf JD, Kitslaar PH, et al. Automated quantification of stenosis severity on 64-slice CT: A comparison with quantitative coronary angiography. JACC Cardiovasc Imaging. 2010;3(7):699–709. doi: 10.1016/j.jcmg.2010.01.010. [DOI] [PubMed] [Google Scholar]
- 3.Nakazato R, Shalev A, Doh JH, et al. Aggregate plaque volume by coronary computed tomography angiography is superior and incremental to luminal narrowing for diagnosis of ischemic lesions of intermediate stenosis severity. J Am Coll Cardiol. 2013;62(5):460–467. doi: 10.1016/j.jacc.2013.04.062. [DOI] [PubMed] [Google Scholar]
- 4.Gould KL. Does coronary flow trump coronary anatomy? JACC: Cardiovascular Imaging. 2009;2(8):1009–1023. doi: 10.1016/j.jcmg.2009.06.004. [DOI] [PubMed] [Google Scholar]
- 5.Gould KL, Johnson NP, Bateman TM, et al. Anatomic versus physiologic assessment of coronary artery disease. role of coronary flow reserve, fractional flow reserve, and positron emission tomography imaging in revascularization decision-making. J Am Coll Cardiol. 2013;62(18):1639–1653. doi: 10.1016/j.jacc.2013.07.076. [DOI] [PubMed] [Google Scholar]
- 6.Vavere AL, Simon GG, George RT, et al. Diagnostic performance of combined noninvasive coronary angiography and myocardial perfusion imaging using 320 row detector computed tomography: Design and implementation of the CORE320 multicenter, multinational diagnostic study. J Cardiovasc Comput Tomogr. 2011;5(6):370–381. doi: 10.1016/j.jcct.2011.11.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Rochitte CE, George RT, Chen MY, et al. Computed tomography angiography and perfusion to assess coronary artery stenosis causing perfusion defects by single photon emission computed tomography: The CORE320 study. Eur Heart J. 2013 doi: 10.1093/eurheartj/eht488. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.George RT, Arbab-Zadeh A, Cerci RJ, et al. Diagnostic performance of combined noninvasive coronary angiography and myocardial perfusion imaging using 320-MDCT: The CT angiography and perfusion methods of the CORE320 multicenter multinational diagnostic study. AJR Am J Roentgenol. 2011;197(4):829–837. doi: 10.2214/AJR.10.5689. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.de Graaf MA, Broersen A, Kitslaar PH, et al. Automatic quantification and characterization of coronary atherosclerosis with computed tomography coronary angiography: Cross-correlation with intravascular ultrasound virtual histology. Int J Cardiovasc Imaging. 2013;29(5):1177–1190. doi: 10.1007/s10554-013-0194-x. [DOI] [PubMed] [Google Scholar]
- 10.de Graaf MA, Broersen A, Ahmed W, et al. Feasibility of an automated quantitative computed tomography angiography-derived risk score for risk stratification of patients with suspected coronary artery disease. Am J Cardiol. 2014;113(12):1947–1955. doi: 10.1016/j.amjcard.2014.03.034. [DOI] [PubMed] [Google Scholar]
- 11.Papadopoulou SL, Neefjes LA, Garcia-Garcia HM, et al. Natural history of coronary atherosclerosis by multislice computed tomography. JACC Cardiovasc Imaging. 2012;5(3 Suppl):S28–37. doi: 10.1016/j.jcmg.2012.01.009. [DOI] [PubMed] [Google Scholar]
- 12.Cerci RJ, Arbab-Zadeh A, George RT, et al. Aligning coronary anatomy and myocardial perfusion territories: An algorithm for the CORE320 multicenter study. Circ Cardiovasc Imaging. 2012;5(5):587–595. doi: 10.1161/CIRCIMAGING.111.970608. [DOI] [PubMed] [Google Scholar]
- 13.Hachamovitch R, Johnson JR, Hlatky MA, et al. The study of myocardial perfusion and coronary anatomy imaging roles in CAD (SPARC): Design, rationale, and baseline patient characteristics of a prospective, multicenter observational registry comparing PET, SPECT, and CTA for resource utilization and clinical outcomes. J Nucl Cardiol. 2009;16(6):935–948. doi: 10.1007/s12350-009-9140-7. [DOI] [PubMed] [Google Scholar]
- 14.Herzog C, Kerl JM, De Rosa S, et al. Influence of observer experience and training on proficiency in coronary CT angiography interpretation. Eur J Radiol. 2013;82(8):1240–1247. doi: 10.1016/j.ejrad.2013.02.037. [DOI] [PubMed] [Google Scholar]
- 15.Papadopoulou SL, Garcia-Garcia HM, Rossi A, et al. Reproducibility of computed tomography angiography data analysis using semiautomated plaque quantification software: Implications for the design of longitudinal studies. Int J Cardiovasc Imaging. 2013;29(5):1095–1104. doi: 10.1007/s10554-012-0167-5. [DOI] [PubMed] [Google Scholar]
- 16.Min JK, Leipsic J, Pencina MJ, et al. Diagnostic accuracy of fractional flow reserve from anatomic CT angiography. JAMA. 2012;308(12):1237–1245. doi: 10.1001/2012.jama.11274. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Arbab-Zadeh A, Nakano M, Virmani R, Fuster V. Acute coronary events. Circulation. 2012;125(9):1147–1156. doi: 10.1161/CIRCULATIONAHA.111.047431. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Voros S, Rinehart S, Qian Z, et al. Prospective validation of standardized, 3-dimensional, quantitative coronary computed tomographic plaque measurements using radiofrequency backscatter intravascular ultrasound as reference standard in intermediate coronary arterial lesions: Results from the ATLANTA (assessment of tissue characteristics, lesion morphology, and hemodynamics by angiography with fractional flow reserve, intravascular ultrasound and virtual histology, and noninvasive computed tomography in atherosclerotic plaques) I study. JACC Cardiovasc Interv. 2011;4(2):198–208. doi: 10.1016/j.jcin.2010.10.008. [DOI] [PubMed] [Google Scholar]
- 19.Arbab-Zadeh A. Fractional flow reserve guided percutaneous coronary intervention is not a valid concept. Circulation. 2014;129(18):1871–1878. doi: 10.1161/CIRCULATIONAHA.113.003583. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Gould KL, Johnson NP, Bateman TM, et al. Anatomic vs. physiologic assessment of coronary artery disease. Role of coronary flow reserve, frational flow reserve, and positron emission tomorgraphy imaging in revascularization decision-making. JACC. 2013;62:1639–53. doi: 10.1016/j.jacc.2013.07.076. [DOI] [PubMed] [Google Scholar]
- 21.Rahsepar A, Arbab-Zadeh A. Cardiac CT vs. stress testing in patients with suspected coronary artery disease: Review and expert recommendations. Curr Cardiovasc Imag Rep. 2015;8:29–40. doi: 10.1007/s12410-015-9344-y. [DOI] [PMC free article] [PubMed] [Google Scholar]