Abstract

Please see the Editorial Comment by Ayaz Aghayev discussing this article.
BACKGROUND. Deep learning frameworks have been applied to interpretation of coronary CTA performed for coronary artery disease (CAD) evaluation.
OBJECTIVE. The purpose of our study was to compare the diagnostic performance of myocardial perfusion imaging (MPI) and coronary CTA with artificial intelligence quantitative CT (AI-QCT) interpretation for detection of obstructive CAD on invasive angiography and to assess the downstream impact of including coronary CTA with AI-QCT in diagnostic algorithms.
METHODS. This study entailed a retrospective post hoc analysis of the derivation cohort of the prospective 23-center Computed Tomographic Evaluation of Atherosclerotic Determinants of Myocardial Ischemia (CREDENCE) trial. The study included 301 patients (88 women and 213 men; mean age, 64.4 ± 10.2 [SD] years) recruited from May 2014 to May 2017 with stable symptoms of myocardial ischemia referred for nonemergent invasive angiography. Patients underwent coronary CTA and MPI before angiography with quantitative coronary angiography (QCA) measurements and fractional flow reserve (FFR). CTA examinations were analyzed using an FDA-cleared cloud-based software platform that performs AI-QCT for stenosis determination. Diagnostic performance was evaluated. Diagnostic algorithms were compared.
RESULTS. Among 102 patients with no ischemia on MPI, AI-QCT identified obstructive (≥ 50%) stenosis in 54% of patients, including severe (≥ 70%) stenosis in 20%. Among 199 patients with ischemia on MPI, AI-QCT identified nonobstructive (1–49%) stenosis in 23%. AI-QCT had significantly higher AUC (all p < .001) than MPI for predicting ≥ 50% stenosis by QCA (0.88 vs 0.66), ≥ 70% stenosis by QCA (0.92 vs 0.81), and FFR < 0.80 (0.90 vs 0.71). An AI-QCT result of ≥ 50% stenosis and ischemia on stress MPI had sensitivity of 95% versus 74% and specificity of 63% versus 43% for detecting ≥ 50% stenosis by QCA measurement. Compared with performing MPI in all patients and those showing ischemia undergoing invasive angiography, a scenario of performing coronary CTA with AIQCT in all patients and those showing ≥ 70% stenosis undergoing invasive angiography would reduce invasive angiography utilization by 39%; a scenario of performing MPI in all patients and those showing ischemia undergoing coronary CTA with AI-QCT and those with ≥ 70% stenosis on AI-QCT undergoing invasive angiography would reduce invasive angiography utilization by 49%.
CONCLUSION. Coronary CTA with AI-QCT had higher diagnostic performance than MPI for detecting obstructive CAD.
CLINICAL IMPACT. A diagnostic algorithm incorporating AI-QCT could substantially reduce unnecessary downstream invasive testing and costs.
TRIAL REGISTRATION. Clinicaltrials.gov NCT02173275

HIGHLIGHTS

Key Finding
Coronary CTA with AI-QCT had higher AUC (all p < .001) than MPI for predicting ≥ 50% stenosis by QCA (0.88 vs 0.66) and FFR < 0.80 (0.90 vs 0.71). AI-QCT ≥ 50% and ischemia on MPI had sensitivity of 95% versus 74% and specificity of 63% versus 43% for detecting ≥ 50% stenosis by QCA.
Importance
A diagnostic algorithm incorporating coronary CTA with AI-QCT could allow substantial reductions in downstream invasive testing and costs.
The diagnostic evaluation of patients with stable chest pain has largely relied on stress myocardial perfusion imaging (MPI) and other functional noninvasive imaging tests for assessment of inducible myocardial ischemia [1, 2]. MPI, including SPECT and PET as well as stress perfusion cardiac MRI, may identify significant coronary artery disease (CAD) based on the observed extent and severity of inducible ischemia [3]. The presence and severity of inducible abnormalities on these tests are the basis for clinical decision making under an ischemia-guided management strategy that may include invasive coronary angiography (ICA) with fractional flow reserve (FFR) and coronary revascularization [3].
Among the noninvasive stress imaging modalities, MPI is the most common, comprising approximately 90% of the more than 10 million stress imaging tests performed annually in the United States [4]. However, MPI has limited performance in ischemia detection. In the prospective Clinical Evaluation of Magnetic Resonance Imaging in Coronary Heart Disease (CE-MARC) trial, SPECT MPI had sensitivity of 66.5%, specificity of 82.6%, PPV of 71.4%, and NPV of 79.1% [5]. In a study that adjusted the performance of SPECT MPI for referral bias using two different formulas, the test yielded a sensitivity of 65–67% and a specificity of 67–75% [6]. Also, in the International Study of Comparative Health Effectiveness With Medical and Invasive Approaches (ISCHEMIA) trial, approximately 20% of patients with moderate or severe ischemia by MPI were found to have nonobstructive CAD at angiography [3].
Coronary CTA is a noninvasive anatomic imaging method with sensitivity of 95–99% for obstructive CAD [7, 8] and is now recognized as a first-line test for the identification and exclusion of obstructive CAD in multiple recent guidelines, including those from the American College of Cardiology (ACC)/American Heart Association (AHA) Joint Committee [9] and the Society of Cardiovascular Computed Tomography (SCCT) [10]. Multicenter studies have consistently shown that the high diagnostic performance of coronary CTA allows reduced downstream testing and improved clinical outcomes [7, 11]. These data have advanced the concept of coronary CTA as a gatekeeper test that more precisely selects individuals to undergo ICA. A recent multicenter clinical trial found that quantitative coronary CTA was superior to conventional stress testing for the diagnosis of vessel-specific FFR, which may translate to further improvements in outcomes [12].
Application of deep learning frameworks to coronary CTA interpretation allows enhanced automation, accuracy, and reliability of evaluation for CAD compared with artificial intelligence (AI) and machine learning and also enables atherosclerosis imaging and quantitative CT [13, 14]. Given its high diagnostic performance, artificial intelligence quantitative CT (AI-QCT) may reduce both overestimation of stenosis severity and false-negative findings compared with MPI. Beyond conventional measures of coronary stenosis severity, AI-QCT enables whole-heart 3D volumetric evaluation of vascular morphology with quantitative assessment of atherosclerotic plaque burden and type; such variables have been shown to be the strongest predictors for future major adverse cardiovascular events [11, 12, 15].
The purpose of this study was to compare the diagnostic performance of stress MPI and coronary CTA with AI-QCT interpretation for the detection of obstructive CAD on invasive angiography as well as to assess the downstream impact of including coronary CTA with AI-QCT in diagnostic algorithms.

Methods

Patients

This study entailed a retrospective post hoc analysis of the derivation cohort of the Computed Tomographic Evaluation of Atherosclerotic Determinants of Myocardial Ischemia (CREDENCE) trial (Clinicaltrials.gov NCT02173275). The study was HIPAA-compliant. The institutional review board of each enrolling site approved the study protocol, and all patients provided written informed consent.
The CREDENCE trial was a prospective multicenter diagnostic derivation-validation controlled clinical trial that recruited patients from May 2014 to May 2017 with stable signs and symptoms suggestive of myocardial ischemia and without prior diagnosis of CAD [12]. Information regarding the 23 participating institutions and the enrolling investigators is provided in Table S1 (available in the online supplement). A prior article was dedicated to presenting the rationale and design of the CREDENCE trial [16]. Trial eligibility required referral for nonemergent ICA based on the ACC/AHA clinical practice guidelines for stable ischemic heart disease [9]. Enrolled patients underwent coronary CTA and stress MPI followed by quantitative ICA with FFR measurements as the reference test within 60 days. All imaging tests were interpreted by physicians at an imaging core laboratory who were blinded to clinical information and the results of the other tests [17]. Patient characteristics were prospectively gathered at the time of coronary CTA examinations.
The CREDENCE trial included 612 patients: 307 in the derivation cohort and 305 in the validation cohort. The 307 patients in the derivation cohort were initially selected for inclusion in the current study. Six of these patients were then excluded because of corrupted media (n = 4) or the absence of evaluation by stress MPI (n = 2), resulting in a final sample for the current study of 301 patients (88 [29%] women and 213 [71%] men; mean age, 64.4 ± 10.2 [SD] years). Figure 1 shows the flow of patient selection for the current study.
Fig. 1 —Flowchart shows patient selection. CREDENCE = Computed Tomographic Evaluation of Atherosclerotic Determinants of Myocardial Ischemia, AI-QCT = artificial intelligence quantitative CT.

Myocardial Perfusion Imaging

Patients underwent rest and stress MPI by SPECT/CT (n = 229 [76%]), PET (n = 24 [8%]), or stress perfusion cardiac MRI (n = 48 [16%]). Rest and stress MPI was performed in accordance with the guidelines of the American Society of Nuclear Cardiology or Society for Cardiovascular Magnetic Resonance [18, 19]. Each MPI examination was interpreted in the core laboratory by one of three cardiologists with 10–15 years of posttraining experience (P.K., R.S.D., G.P.), including one (P.K.) with expertise in both MPI interpretation and interventional cardiology. Segmental scores were aggregated per patient and vascular territory. Perfusion in each of the 17 coronary segments in the AHA/ACC model [20] was graded semiquantitatively as follows: 0, normal; 1, equivocal or mild reduction; 2, moderate reduction; 3, severe reduction; or 4, absent uptake. Segmental scores were summed for the stress scans, yielding the summed stress score (SSS). The SSS was classified as follows: 0, no ischemia; 1–8, minimal or mild ischemia; 9–13, moderate ischemia; or greater than 13, severe ischemia.

Coronary CTA Acquisition

Coronary CTA was acquired using a single-source or dual-source CT scanner with at least 64 detector rows. Coronary CTA was acquired in accordance with SCCT guidelines [21]. Patients were administered nitroglycerin immediately before the acquisition.

Artificial Intelligence–Based Interpretation of Coronary CTA

AI-QCT was performed using an FDA-cleared cloud-based commercially available software platform (Cleerly Labs, Cleerly) that provides coronary CTA analysis. The software uses a series of validated convolutional neural network models for quantitative image quality assessment, coronary segmentation and labeling, lumen and vessel wall determination, vascular morphology measurements, atherosclerotic plaque quantification and characterization, and stenosis determination [13, 14]. As mandated by the FDA, the AI-QCT analysis in each patient underwent a quality review in the core laboratory by a technologist trained in cardiac CT. Two multicenter clinical trials have reported validation of AI-QCT [13, 14].
The analysis included coronary segments having a diameter of 2 mm or greater based on a modified 18-segment SCCT model [22]. Segments were assessed for the presence versus absence of atherosclerosis, defined as tissue structures measuring greater than 1 mm3 within the coronary artery wall, and were differentiated from surrounding epicardial tissue, epicardial fat, or the vessel lumen itself. The presence of stenosis was determined using a normal proximal reference vessel on a cross-sectional slice. Each lesion's start and end were identified, and the percent diameter stenosis was automatically calculated from the cross-sectional slice that showed the greatest absolute narrowing. Obstructive stenosis was defined at thresholds of both ≥ 50% and ≥ 70%.
All studies were analyzable by AI-QCT and included in the study results. Among the total of 171,195 mm of vessel length evaluated in the entire cohort, a total of 1861 mm (1.09%) of vessel length was excluded.

Quantitative Coronary Angiography Measurements

Quantitative coronary angiography (QCA) measurements were performed in the core laboratory by one of four cardiologists with level 3 training (P.K., M.J.B., H.C., P.G.). QCA measurements were obtained in two orthogonal views on a per-lesion basis for every lesion visually showing ≥ 30% diameter stenosis in vessels with a reference vessel diameter of 2.0 mm or greater. QCA measurements were obtained in each coronary vessel territory, and AI-based stenosis evaluation was performed for each present coronary segment using the 18-segment SCCT coronary segmentation model. When the QCA measurement for a vessel territory was 0%, the 0% measurement was applied to each segment in the territory. The maximum QCA measurement and AI-based diameter stenosis were calculated across segments for each of the 10 vessels. When the QCA measurement was obtained in just one segment in the vessel, the obtained measurement was applied to the vessel. Per-vessel territory and per-patient results were obtained in a similar manner. Stenosis of ≥ 50% was interpreted as representing obstructive CAD, and stenosis of ≥ 70% was interpreted as representing severe obstructive CAD.

Invasive Fractional Flow Reserve

All major coronary arteries or branches (diameter ≥ 2.0 mm) containing a lesion with between 40% and 90% stenosis were interrogated by FFR during intracoronary (150 μg) or IV (140 mcg/kg/min) adenosine infusion to achieve maximal hyperemia, as defined in the primary CREDENCE protocol.

Sequential Testing Models and Costs

A model of sequential testing was established to evaluate the estimated rate of downstream invasive angiography after MPI in the presence of ischemia versus an “AI-QCT first” approach set at a ≥ 70% obstructive threshold. The costs of coronary CTA, stress MPI, and invasive angiography were based on information from the Hospital Outpatient Prospective Payment System (HOPPS) [23, 24] and the Medicare Physician Fee Schedule [25], both accessed January 25, 2022. Specifically, the model assumed a cost per examination of $1413 for stress MPI, $298 for coronary CTA, and $3205 for invasive angiography. The cost of AI-QCT is not currently listed by HOPPS and was thus set at the vendor (Cleerly) list price of $1500 per examination. The cost of coronary CTA with AI-QCT interpretation was derived as the sum of the two component costs ($1798).
Costs were modeled based on four scenarios: scenario A (baseline), all patients underwent MPI, and those showing any ischemia then underwent diagnostic invasive angiography; scenario B, all patients underwent coronary CTA and AI-QCT, and those with ≥ 70% stenosis then underwent invasive angiography; scenario C, all patients underwent MPI, those showing any ischemia then underwent coronary CTA and AI-QCT, and those with ≥ 70% stenosis on coronary CTA and AI-QCT then underwent invasive angiography; and scenario D, all patients underwent coronary CTA and AI-QCT, and those with ≥ 70% stenosis then underwent invasive angiography, whereas those with 50–69% stenosis then underwent MPI and ultimately invasive angiography if MPI showed any ischemia.

Statistical Analysis

Continuous data were reported as mean ± SD, and categoric variables were reported as absolute numbers with percentages. The cross-tabulation of AI-QCT and stress MPI findings was summarized descriptively. The diagnostic performance of AI-QCT–based diameter stenosis at both ≥ 50% and ≥ 70% thresholds and of stress MPI (stratified as no ischemia vs any ischemia) was evaluated by calculating the sensitivity, specificity, PPV, and NPV on a per-patient basis and assessing for endpoints of QCA ≥ 50%, QCA ≥ 70%, and FFR < 0.80. ROC curves were generated to compare the diagnostic performance of AI-QCT (in terms of the percent stenosis) and stress MPI (in terms of the score from 0 to 4) for these same three endpoints. The AUCs were compared for statistically significant differences by using the chi-square test to evaluate the contrast matrix. The mean cost per patient and the number of patients who underwent invasive angiography were compared between the baseline testing scenario and the three other sequential testing scenarios; p values < .05 were considered statistically significant. Statistical analyses were performed using SAS software (version 9.4, SAS Institute).

Results

Table 1 depicts the demographics of the study sample. In terms of cardiovascular risk factors, 64% of patients had hyper-tension, 45% had dyslipidemia, and 49% had current or recent smoking history. A total of 36% of patients had typical angina, 16% had atypical angina, and 18% had dyspnea. Based on coronary angiography as the reference standard, less than 1% (1/301) of patients had no CAD, 32% (94/301) had nonobstructive CAD (< 50% stenosis), 35% (103/301) had one-vessel obstructive CAD, 17% (51/301) had two-vessel obstructive CAD, and 16% (46/301) had three-vessel obstructive CAD or left main obstructive CAD; 41% (122/301) had severe obstructive CAD (≥ 70% stenosis).
TABLE 1: Characteristics of Study Patients
CharacteristicValue (n = 301)
Age (y), mean ± SD64.4 ± 10.2
Sex 
 Female88 (29)
 Male213 (71)
Hypertension193 (64)
Dyslipidemia134 (45)
Diabetes93 (31)
Family history of CAD58 (19)
Current smoker or history of smoking ≤ 1 y146 (49)
Chest pain-related symptoms 
 Typical angina109 (36)
 Atypical angina48 (16)
 Noncardiac chest pain40 (13)
 No symptoms with abnormal diagnostic test104 (35)
Other symptoms 
 Dyspnea53 (18)
 Palpitations12 (4)
 Dizziness or syncope2 (< 1)
 Arrhythmia4 (1)
Presence of CAD and obstructive (≥ 50% stenosis) CAD on invasive angiography (n = 295) 
 No CAD1 (< 1)
 Nonobstructive CAD94 (32)
 One-vessel obstructive CAD103 (35)
 Two-vessel obstructive CAD51 (17)
 Three-vessel obstructive CAD or LM obstructive CAD46 (16)

Note—Unless otherwise indicated, data represent number of patients with percentage in parentheses. Some percentages may not total 100 because of rounding. CAD = coronary artery disease, LM = left main coronary artery.

Cross-Tabulation of AI-QCT and Stress Myocardial Perfusion Imaging Findings

Table 2 and Figure 2 show the cross-tabulation of AI-QCT and stress MPI findings. Among the 102 patients with no ischemia on stress MPI, AI-QCT identified nonobstructive (1–49%) stenosis in 46% (47/102) and obstructive (≥ 50%) stenosis in 54% (55/102), including severe (≥ 70%) stenosis in 20% (20/102). Among the 199 patients with ischemia on stress MPI, 50% (100/199) had minimal or mild ischemia on MPI, and 50% (99/199) had moderate or severe ischemia on MPI. Among the patients with ischemia on stress MPI, AI-QCT identified nonobstructive (1–49%) stenosis in 24% (47/199), 50–69% stenosis in 25% (49/199), and severe (≥ 70%) obstructive stenosis in 51% (102/199). Among the 99 patients with moderate or severe ischemia on stress MPI, AI-QCT identified nonobstructive (1–49%) stenosis in 10% (10/99), 50–69% stenosis in 14% (14/99), and severe (≥ 70%) obstructive stenosis in 76% (75/99).
TABLE 2: Cross-Tabulation of Findings From Stress MPI and AI-QCT
AI-QCTStress MPIa
No Ischemia (n = 102)Any Ischemia (n = 199)Minimal or Mild Ischemia (n = 100)Moderate Ischemia (n = 42)Severe Ischemia (n = 57)
0% Stenosis (n = 1)0 (0)1 (< 1)1 (1)0 (0)0 (0)
1–49% Stenosis (n = 94)47 (46)47 (24)37 (37)5 (12)5 (9)
50–69% Stenosis (n = 84)35 (34)49 (25)35 (35)7 (17)7 (12)
≥ 70% Stenosis (n = 122)20 (20)102 (51)27 (27)30 (71)45 (79)

Note—Data represent number of patients with percentage in parentheses; some percentages may not total 100 because of rounding. MPI = myocardial perfusion imaging, AI-QCT = artificial intelligence quantitative CT.

a
Ischemia was defined as follows using the summed stress score (SSS): SSS of 0, no ischemia; 1–8, minimal or mild ischemia; 9–13, moderate ischemia; and greater than 13, severe ischemia.
Fig. 2 —Bar graph shows distribution of artificial intelligence quantitative CT (AI-QCT) results for patients with varying stress myocardial perfusion imaging (MPI) results; values inside bars are percentages of patients within MPI-stratified category in whom certain percent of stenosis was identified by AI-QCT. Total of 54% (55/102) of patients with no ischemia on stress MPI showed ≥ 50% stenosis by AI-QCT, 24% (47/199) of patients with any ischemia on stress MPI showed 1–49% stenosis by AI-QCT, and 27% of patients with minimal or mild ischemia on stress MPI showed ≥ 70% stenosis by AI-QCT. Moderate and severe ischemia categories on stress MPI are grouped for illustrative purposes.

Diagnostic Performance of AI-QCT and Stress Myocardial Perfusion Imaging Using Quantitative Coronary Angiography Measurement and Invasive Fractional Flow Reserve as Reference Standard

Table 3 shows the diagnostic performance of AI-QCT and stress MPI for prediction of findings from QCA measurement and invasive FFR. AI-QCT result of ≥ 50% stenosis and ischemia on stress MPI had sensitivity of 95% versus 74%, specificity of 63% versus 43%, PPV of 75% versus 61%, and NPV of 92% versus 59% for detection of ≥ 50% stenosis by QCA measurement. AI-QCT result of ≥ 70% stenosis and ischemia on stress MPI had sensitivity of 95% versus 91%, specificity of 81% versus 44%, PPV of 67% versus 39%, and NPV of 98% versus 92% for detection of ≥ 70% stenosis by QCA measurement. For detection of FFR < 0.80, an AI-QCT result of ≥ 50% stenosis, AI-QCT result of ≥ 70% stenosis, and ischemia on stress MPI had sensitivity of 95%, 70%, and 78%, respectively; specificity of 62%, 93%, and 47%; PPV of 74%, 92%, and 63%; and NPV of 92%, 73%, and 65%.
TABLE 3: Comparison of Diagnostic Performance of AI-QCT and Stress MPI for Coronary Artery Disease Using QCA or Invasive FFR as Reference Standard
Reference Standard, Diagnostic TestSensitivitySpecificityPPVNPV
QCA ≥ 50% stenosis    
 AI-QCT ≥ 50% stenosis95 (155/163)63 (87/138)75 (155/206)92 (87/95)
 Ischemia on MPI74 (121/163)43 (60/138)61 (121/199)59 (60/102)
QCA ≥ 70% stenosis    
 AI-QCT ≥ 70% stenosis95 (82/86)81 (175/215)67 (82/122)98 (175/179)
 Ischemia on MPI91 (78/86)44 (94/215)39 (78/199)92 (94/102)
FFR < 0.80    
 AI-QCT ≥ 50% stenosis95 (153/161)62 (87/140)74 (153/206)92 (87/95)
 AI-QCT ≥ 70% stenosis70 (112/161)93 (130/140)92 (112/122)73 (130/179)
 Ischemia on MPI78 (125/161)47 (66/140)63 (125/199)65 (66/102)

Note—Data are percentages with numerator and denominator in parentheses. AI-QCT = artificial intelligence quantitative CT, MPI = myocardial perfusion imaging, QCA = quantitative coronary angiography, FFR = fractional flow reserve.

In ROC analysis (Fig. 3), AUC was significantly higher for AI-QCT than for stress MPI for prediction of ≥ 50% stenosis by QCA measurement (0.88 [95% CI, 0.84–0.92] vs 0.66 [95% CI, 0.60–0.72], p < .001), prediction of ≥ 70% stenosis by QCA (0.92 [95% CI, 0.90–0.95] vs 0.81 [95% CI, 0.76–0.87], p < .001), and prediction of FFR < 0.80 (0.90 [95% CI, 0.87–0.94] vs 0.71 [95% CI, 0.66–0.77, p < .001). Figure 4 and Figure S1 (available in the online supplement) show representative patients in whom AI-QCT analysis was concordant with invasive angiography and FFR but stress MPI exhibited false-positive or false-negative findings, respectively.
Fig. 3A —ROC curves.
A, ROC curves of artificial intelligence quantitative CT (AI-QCT) and stress myocardial perfusion imaging (MPI) results for endpoints of ≥ 50% stenosis on quantitative coronary angiography (QCA) (A), ≥ 70% stenosis on QCA (B), and fractional flow reserve < 0.80 (C). For all three endpoints, AUC was significantly higher for AI-QCT than for stress MPI (all p < .001).
Fig. 3B —ROC curves.
B, ROC curves of artificial intelligence quantitative CT (AI-QCT) and stress myocardial perfusion imaging (MPI) results for endpoints of ≥ 50% stenosis on quantitative coronary angiography (QCA) (A), ≥ 70% stenosis on QCA (B), and fractional flow reserve < 0.80 (C). For all three endpoints, AUC was significantly higher for AI-QCT than for stress MPI (all p < .001).
Fig. 3C —ROC curves.
C, ROC curves of artificial intelligence quantitative CT (AI-QCT) and stress myocardial perfusion imaging (MPI) results for endpoints of ≥ 50% stenosis on quantitative coronary angiography (QCA) (A), ≥ 70% stenosis on QCA (B), and fractional flow reserve < 0.80 (C). For all three endpoints, AUC was significantly higher for AI-QCT than for stress MPI (all p < .001).
Fig. 4A —55-year-old woman who presented with chest pain.
A, Curved (A) and straightened (B) reformatted CTA images of right coronary artery (RCA). Artificial intelligence quantitative CT (AI-QCT) additions to B show lumen boundaries (yellow lines), and pink lines demarcate border between lumen and plaque. AI-QCT analysis (Cleerly Labs, Cleerly) of these images showed 9% stenosis of RCA.
Fig. 4B —55-year-old woman who presented with chest pain.
B, Curved (A) and straightened (B) reformatted CTA images of right coronary artery (RCA). Artificial intelligence quantitative CT (AI-QCT) additions to B show lumen boundaries (yellow lines), and pink lines demarcate border between lumen and plaque. AI-QCT analysis (Cleerly Labs, Cleerly) of these images showed 9% stenosis of RCA.
Fig. 4C —55-year-old woman who presented with chest pain.
C, Curved (C) and straightened (D) reformatted CTA images of left main artery (LM) and left anterior descending artery (LAD) show plaque (arrows) in LM and LAD. AI-QCT additions to D show lumen boundaries (yellow lines) and calcified (blue shaded area) and noncalcified (yellow shaded area) plaque, and pink lines demarcate border between lumen and plaque. AI-QCT analysis of these images showed 41% stenosis of LM and LAD.
Fig. 4D —55-year-old woman who presented with chest pain.
D, Curved (C) and straightened (D) reformatted CTA images of left main artery (LM) and left anterior descending artery (LAD) show plaque (arrows) in LM and LAD. AI-QCT additions to D show lumen boundaries (yellow lines) and calcified (blue shaded area) and noncalcified (yellow shaded area) plaque, and pink lines demarcate border between lumen and plaque. AI-QCT analysis of these images showed 41% stenosis of LM and LAD.
Fig. 4E —55-year-old woman who presented with chest pain.
E, Curved (E) and straightened (F) reformatted CTA images of left circumflex artery (LCx) show LCx. AI-QCT additions to F show lumen boundaries (yellow lines) and calcified (blue shaded area) and noncalcified (yellow shaded area) plaque, and pink lines demarcate border between lumen and plaque. AI-QCT analysis showed 30% stenosis of LCx. In this patient, AI-QCT results of 9% stenosis of RCA, 41% stenosis of LM and LAD, and 30% stenosis of LCx are consistent with nonobstructive stenosis.
Fig. 4F —55-year-old woman who presented with chest pain.
F, Curved (E) and straightened (F) reformatted CTA images of left circumflex artery (LCx) show LCx. AI-QCT additions to F show lumen boundaries (yellow lines) and calcified (blue shaded area) and noncalcified (yellow shaded area) plaque, and pink lines demarcate border between lumen and plaque. AI-QCT analysis showed 30% stenosis of LCx. In this patient, AI-QCT results of 9% stenosis of RCA, 41% stenosis of LM and LAD, and 30% stenosis of LCx are consistent with nonobstructive stenosis.
Fig. 4G —55-year-old woman who presented with chest pain.
G, Multiplane perfusion (G) and polar (H) maps from stress myocardial perfusion imaging (MPI) show anterior ischemia. In H, numbers in left images show semiquantitative grades of perfusion by coronary segment. Segmental scores were summed for stress scans, yielding summed stress score (SSS). In this case, SSS was 10, which is consistent with severe ischemia. In H, ANT = anterior, LAT = lateral, INF = inferior, SEPT = septal.
Fig. 4H —55-year-old woman who presented with chest pain.
H, Multiplane perfusion (G) and polar (H) maps from stress myocardial perfusion imaging (MPI) show anterior ischemia. In H, numbers in left images show semiquantitative grades of perfusion by coronary segment. Segmental scores were summed for stress scans, yielding summed stress score (SSS). In this case, SSS was 10, which is consistent with severe ischemia. In H, ANT = anterior, LAT = lateral, INF = inferior, SEPT = septal.
Fig. 4I —55-year-old woman who presented with chest pain.
I, Invasive Invasive angiography shows nonobstructive coronary artery disease involving LAD (arrows), with invasive fractional flow reserve (FFR) greater than 0.80. Thus, AI-QCT was concordant with invasive angiography and FFR, whereas stress MPI provided false-positive results.

Sequential Testing Models

Table 4 summarizes the comparisons of four testing models in terms of mean cost per patient, number of patients who underwent invasive angiography, and changes from the baseline testing model. The baseline scenario (scenario A), in which stress MPI was initially performed in all patients and those showing ischemia then underwent invasive angiography, had a mean per-patient cost of $3532. In the second scenario (scenario B), in which patients underwent coronary CTA and AI-QCT and those showing ≥ 70% stenosis underwent invasive angiography, had a mean per-patient cost of $3097 (14% reduction vs baseline scenario); the number of patients who underwent invasive angiography decreased from 199 to 122 (39% reduction). In the third scenario (scenario C), in which patients underwent MPI, those showing any ischemia underwent coronary CTA and AI-QCT, and those with ≥ 70% stenosis on coronary CTA and AI-QCT underwent invasive angiography, had a mean per-patient cost of $3332 (6% reduction); the number of patients who underwent invasive angiography decreased from 199 to 102 (49% reduction). In this scenario, 97 of 199 patients with ischemia on MRI would have been able to defer invasive angiography (48 with 1–49% stenosis on AI-QCT and 49 with 50–69% stenosis on AI-QCT). In the fourth scenario (scenario D), in which patients underwent coronary CTA and AIQCT and those with ≥ 70% stenosis then underwent invasive angiography, while those with 50–69% stenosis underwent MPI and ultimately invasive angiography if MPI showed any ischemia, had a mean per-patient cost of $4013 (14% reduction); the number of patients who underwent invasive angiography decreased from 199 to 171 (14% reduction).
TABLE 4: Downstream Impact of Various Sequential Testing Strategies
ScenarioNo. of PatientsTotal Costs($)aMean Cost per Patient ($)Change From Baseline Scenario (%)
A. Baseline: MPI first, ICA if MPI shows ischemia    
 Stress MPI301425,3131413 
 ICA199637,7953205
 All tests 1,063,1083532
B. AI-QCT first, ICA if AI-QCT shows ≥ 70% stenosis    
 AI-QCT301541,1981798 
 ICA122391,0103205 
 All tests 932,2083097−14
C. MPI first, AI-QCT if MPI shows ischemia, ICA if AI-QCT shows ≥ 70% stenosis    
 Stress MPI301425,3131413 
 AI-QCT199357,8021798 
 ICA102326,9103205 
 All tests 1,110,0253688−6
D. AI-QCT first, MPI if AI-QCT shows 50–69% stenosis, ICA if MPI shows ischemia or if AI-QCT shows ≥ 70% stenosis    
 AI-QCT301541,1981798 
 Stress MPI84118,6921413 
 ICA171548,0553205 
 All tests 1,207,9454013−14

Note—Dash (—) indicates not applicable. MPI = myocardial perfusion imaging, ICA = invasive coronary angiography, AI-QCT = artificial intelligence quantitative CT.

a
Costs are based on 2022 Hospital Outpatient Prospective Payment System [24] and Medicare Physician Fee Schedule [25] Final Rules, accessed January 31, 2022, aside from use of vendor list price for AI-QCT (Cleerly Labs, Cleerly). Model assumes cost per examination for stress MPI was $1413; for AI-QCT, $1798 ($298 for coronary CTA + $1500 for artificial intelligence component); and for ICA, $3205.

Discussion

In this 23-center study of patients referred to nonemergent ICA based on the ACC/AHA clinical practice guideline for stable ischemic heart disease, AI-QCT had higher sensitivity and specificity than stress MPI for obstructive CAD using invasive angiography as the reference standard, allowing substantial reductions in downstream invasive testing and costs. The higher diagnostic performance of AI-QCT than of stress MPI was observed whether using an endpoint of ≥ 50% stenosis by QCA measurement, ≥ 70% stenosis by QCT measurement, or invasive FFR < 0.80. An “AI-QCT first” approach and using a ≥ 70% threshold on AI-QCT for invasive angiography (scenario B) yielded a 39% reduction in downstream invasive angiography utilization and an estimated 14% cost savings; an approach incorporating AI-QCT after positive stress MPI and likewise using a ≥ 70% threshold on AI-QCT for invasive angiography (scenario C) yielded a 49% reduction in invasive angiography utilization and an estimated cost savings of 6%. Finally, among patients with no ischemia on stress MPI, AIQCT identified obstructive (≥ 50%) stenosis in 54% and severe (≥ 70%) stenosis in 20%. These patients with obstructive CAD on AIQCT risk receiving an insufficient treatment regimen if undergoing only nuclear stress testing.
MPI is a widely performed noninvasive stress imaging test. However, MPI underestimated and overestimated vessel-specific ischemia in 36% and 22%, respectively, of patients with multivessel CAD using FFR as the reference standard [26]. This suboptimal real-world accuracy of MPI was noted in the recent International Study of Comparative Health Effectiveness With Medical and Invasive Approaches (ISCHEMIA) trial, in which approximately 20% of patients with moderate or severe ischemia by MPI were found to have nonobstructive CAD at invasive angiography [3]. Such data evoke concerns for the clinical ability of MPI to effectively identify coronary lesions that would benefit from revascularization [26]. In the current study, stress MPI had low specificity of 44% using ≥ 70% stenosis by QCA measurement as reference standard and low specificity of 47% using FFR < 0.80 as reference standard compared with specificity for AI-QCT of 81% and 93%, respectively.
Our data suggest that AI-guided coronary CTA analysis would enable substantial reduction in downstream testing by reliably excluding obstructive CAD at predefined stenosis thresholds, thereby avoiding unnecessary invasive angiography in many patients. It is estimated that over 1 million cardiac catheterizations to evaluate for CAD are performed in the United States annually [27]. However, obstructive CAD is found on invasive angiography in only approximately 40% of patients despite prior noninvasive stress testing being performed in approximately 60–85% of all patients [27]. A prior study suggested that the use of coronary CTA, rather than nuclear stress tests, to select patients for invasive angiography could increase the diagnostic yield of invasive angiography and reduce the number of unnecessary invasive tests [28]. The use of AI-QCT may have further benefit relative to coronary CTA without AI, providing specific stenosis estimates to guide determinations of the necessity of invasive angiography based on consistent thresholds (e.g., ≥ 70%).
In addition to the ability of AI-guided coronary CTA analysis to facilitate correct CAD diagnoses [29], the method also offers the opportunity to substantially reduce the amount of radiation received by patients with suspected CAD. The use of radiation for medical treatment and imaging in the United States has been increasing for decades [30]. A single nuclear MPI examination results in a radiation exposure equivalent of 150 chest radiographs (9 mSv of radiation), and nuclear MPI is estimated to account for over 10% of the U.S. population's cumulative radiation from all sources excluding radiotherapy [30]. Coronary CTA is associated with substantially lower radiation exposure to the patient in comparison with nuclear MPI. The recent global Prospective Multicenter Registry on Radiation Dose Estimates of Cardiac CT Angiography in Daily Practice in 2017 (PROTECTION-VI) registry found that with the application of low tube potential, coronary CTA delivered a mean dose of 2–4 mSv [31]. Modern scanning protocols allow coronary CTA to be performed at less than 1 mSv [32].
This study has limitations. It was a retrospective post hoc analysis of data from a prospective trial. The findings may have limited generalizability given that 71% of the patients were men. In addition, the various sequential testing models were explored in a simulated fashion, and actual clinical outcomes resulting from these models are unknown. Also, the models assume that all positive diagnostic tests results would be referred to invasive angiography and that invasive angiography would have perfect sensitivity and specificity. The models further assume that patients would not undergo tests outside of the simulated pathway. Real-world practice is expected to differ from these assumptions. Finally, AI-QCT was not compared with coronary CTA without AI assistance given prior studies validating AI-QCT in such a comparison [13, 14].

Conclusion

In this study, AI-QCT identified obstructive stenoses in 54% of patients with normal stress MPI, and AI-QCT had significantly higher diagnostic performance than stress MPI in detecting obstructive CAD using invasive angiography as the reference standard. The incorporation of AI-QCT in diagnostic paradigms, whether performed first in patients with suspected obstructive CAD or performed after stress MPI shows evidence of ischemia, allows substantial reduction in unnecessary downstream invasive angiography procedures and associated costs. These findings could help inform future approaches to the diagnostic work-up of patients with suspected CAD.

Supplemental Content

File (21_27289_suppl.pdf)

References

1.
Fihn SD, Gardin JM, Abrams J, et al.; American College of Cardiology Foundation; American Heart Association Task Force on Practice Guidelines; American College of Physicians; American Association for Thoracic Surgery; Preventive Cardiovascular Nurses Association; Society for Cardiovascular Angiography and Interventions; Society of Thoracic Surgeons. 2012 ACCF/AHA/ACP/AATS/PCNA/SCAI/STS guideline for the diagnosis and management of patients with stable ischemic heart disease: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines, and the American College of Physicians, American Association for Thoracic Surgery, Preventive Cardiovascular Nurses Association, Society for Cardiovascular Angiography and Interventions, and Society of Thoracic Surgeons. J Am Coll Cardiol 2012; 60:e44–e164
2.
Fihn SD, Blankenship JC, Alexander KP, et al. 2014 ACC/AHA/AATS/PCNA/SCAI/STS focused update of the guideline for the diagnosis and management of patients with stable ischemic heart disease: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines, and the American Association for Thoracic Surgery, Preventive Cardiovascular Nurses Association, Society for Cardiovascular Angiography and Interventions, and Society of Thoracic Surgeons. J Am Coll Cardiol 2014; 64:1929–1949
3.
Maron DJ, Hochman JS, Reynolds HR, et al.; ISCHEMIA Research Group. Initial invasive or conservative strategy for stable coronary disease. N Engl J Med 2020; 382:1395–1407
4.
Shaw L, Kwong RY, Nagel E, et al. Cardiac imaging in the post-ISCHEMIA trial era: a multisociety viewpoint. JACC Cardiovasc Imaging 2020; 13:1815–1833
5.
Greenwood JP, Maredia N, Younger JF, et al. Cardiovascular magnetic resonance and single-photon emission computed tomography for diagnosis of coronary heart disease (CE-MARC): a prospective trial. Lancet 2012; 379:453–460
6.
Miller TD, Hodge DO, Christian TF, Milavetz JJ, Bailey KR, Gibbons RJ. Effects of adjustment for referral bias on the sensitivity and specificity of single photon emission computed tomography for the diagnosis of coronary artery disease. Am J Med 2002; 112:290–297
7.
Budoff M J, Dowe D, Jollis JG, et al. Diagnostic performance of 64-multidetector row coronary computed tomographic angiography for evaluation of coronary artery stenosis in individuals without known coronary artery disease: results from the prospective multicenter ACCURACY (Assessment by Coronary Computed Tomographic Angiography of Individuals Undergoing Invasive Coronary Angiography) trial. J Am Coll Cardiol 2008; 52:1724–1732
8.
Chow BJ, Abraham A, Wells GA, et al. Diagnostic accuracy and impact of computed tomographic coronary angiography on utilization of invasive coronary angiography. Circ Cardiovasc Imaging 2009; 2:16–23
9.
Writing Committee Members; Gulati M, Levy PD, Mukherjee D, et al. 2021 AHA/ACC/ASE/CHEST/SAEM/SCCT/SCMR Guideline for the evaluation and diagnosis of chest pain: a report of the American College of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines. J Am Coll Cardiol 2021; 78:e187–e285
10.
Narula J, Chandrashekhar Y, Ahmadi A, et al. SCCT 2021 expert consensus document on coronary computed tomographic angiography: a report of the Society of Cardiovascular Computed Tomography. J Cardiovasc Comput Tomogr 2021; 15:192–217
11.
SCOT-HEART Investigators; Newby DE, Adamson PD, Berry C, et al. Coronary CT angiography and 5-year risk of myocardial infarction. N Engl J Med 2018; 379:924–933
12.
Stuijfzand WJ, van Rosendael AR, Lin FY, et al.; CREDENCE Investigators. Stress myocardial perfusion imaging vs coronary computed tomographic angiography for diagnosis of invasive vessel-specific coronary physiology: predictive modeling results from the Computed Tomographic Evaluation of Atherosclerotic Determinants of Myocardial Ischemia (CREDENCE) Trial. JAMA Cardiol 2020; 5:1338–1348
13.
Choi AD, Marques H, Kumar V, et al. CT Evaluation by Artificial Intelligence for Atherosclerosis, Stenosis and Vascular Morphology (CLARIFY): a multi-center, international study. J Cardiovasc Comput Tomogr 2021; 15:470–476
14.
Griffin WF, Choi AD, Riess J, et al. AI evaluation of coronary stenosis on CT coronary angiography, comparison with quantitative coronary angiography and fractional flow reserve: a CREDENCE trial substudy. JACC Cardiovasc Imaging 2022 Feb 15 [published online]
15.
Shaw LJ, Blankstein R, Bax JJ, et al. Society of Cardiovascular Computed Tomography/North American Society of Cardiovascular Imaging: expert consensus document on coronary CT imaging of atherosclerotic plaque. J Cardiovasc Comput Tomogr 2021; 15:93–109
16.
Rizvi A, Hartaigh BÓ, Knaapen P, et al. Rationale and design of the CREDENCE trial: Computed Tomographic Evaluation of Atherosclerotic Determinants of Myocardial Ischemia. BMC Cardiovasc Disord 2016; 16:190
17.
Fihn SD, Gardin JM, Abrams J, et al. 2012 ACCF/AHA/ACP/AATS/PCNA/SCAI/STS guideline for the diagnosis and management of patients with stable ischemic heart disease: executive summary: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines, and the American College of Physicians, American Association for Thoracic Surgery, Preventive Cardiovascular Nurses Association, Society for Cardiovascular Angiography and Interventions, and Society of Thoracic Surgeons. J Am Coll Cardiol 2012; 60:2564–2603
18.
Kramer CM, Barkhausen J, Flamm SD, Kim RJ, Nagel E; Society for Cardiovascular Magnetic Resonance Board of Trustees Task Force on Standardized Protocols. Standardized cardiovascular magnetic resonance imaging (CMR) protocols, society for cardiovascular magnetic resonance: board of trustees task force on standardized protocols. J Cardiovasc Magn Reson 2008; 10:35
19.
Henzlova MJ, Duvall WL, Einstein AJ, Travin MI, Verberne HJ. ASNC imaging guidelines for SPECT nuclear cardiology procedures: stress, protocols, and tracers. J Nucl Cardiol 2016; 23:606–639
20.
Cerqueira MD, Weissman NJ, Dilsizian V, et al.; American Heart Association Writing Group on Myocardial Segmentation and Registration for Cardiac Imaging. Standardized myocardial segmentation and nomenclature for tomographic imaging of the heart: a statement for healthcare professionals from the Cardiac Imaging Committee of the Council on Clinical Cardiology of the American Heart Association. Circulation 2002; 105:539–542
21.
Abbara S, Blanke P, Maroules CD, et al. SCCT guidelines for the performance and acquisition of coronary computed tomographic angiography: a report of the society of Cardiovascular Computed Tomography Guidelines Committee—endorsed by the North American Society for Cardiovascular Imaging (NASCI). J Cardiovasc Comput Tomogr 2016; 10:435–449
22.
Leipsic J, Abbara S, Achenbach S, et al. SCCT guidelines for the interpretation and reporting of coronary CT angiography: a report of the Society of Cardiovascular Computed Tomography Guidelines Committee. J Cardiovasc Comput Tomogr 2014; 8:342–358
23.
American Society of Nuclear Cardiology website. Reimbursement & coding procedures. www.asnc.org/coding_reimbursement. Published 2020. Accessed September 30, 2021
24.
Centers for Medicare & Medicaid Services (CMS) website. Hospital outpatient PPS: addendum A and addendum B updates. www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/HospitalOutpatientPPS/Addendum-A-and-Addendum-B-Updates. Published 2022. Accessed January 31, 2022
25.
Centers for Medicare & Medicaid Services (CMS) website. Center for Medicare and Medicaid Services Physician Fee Schedule. www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/PhysicianFeeSched. Accessed January 31, 2022
26.
Beller GA, Ragosta M. Decision making in multivessel coronary disease: the need for physiological lesion assessment. JACC Cardiovasc Interv 2010; 3:315–317
27.
Patel MR, Peterson ED, Dai D, et al. Low diagnostic yield of elective coronary angiography. N Engl J Med 2010; 362:886–895
28.
Gonçalves M, Roque D, de Araújo Gonçalves P, et al. Temporal trends in referral patterns for invasive coronary angiography: a multicenter 10-year analysis. Coron Artery Dis 2021; 32:224–230
29.
Lu MT, Meyersohn NM, Mayrhofer T, et al. Central core laboratory versus site interpretation of coronary CT angiography: agreement and association with cardiovascular events in the PROMISE trial. Radiology 2018; 287:87–95
30.
Mettler FA Jr. Medical radiation exposure in the United States: 2006–2016 trends. Health Phys 2019; 116:126–128
31.
Stocker TJ, Leipsic J, Hadamitzky M, et al. Application of low tube potentials in CCTA: results from the PROTECTION VI study. JACC Cardiovasc Imaging 2020; 13:425–434
32.
Chen MY, Shanbhag SM, Arai AE. Submillisievert median radiation dose for coronary angiography with a second-generation 320-detector row CT scanner in 107 consecutive patients. Radiology 2013; 267:76–85

Information & Authors

Information

Published In

American Journal of Roentgenology
Pages: 407 - 419
PubMed: 35441530

History

Submitted: December 23, 2021
Revision requested: January 5, 2022
Revision received: March 14, 2022
Accepted: March 29, 2022
Version of record online: April 20, 2022

Keywords

  1. artificial intelligence
  2. atherosclerosis
  3. CCTA
  4. coronary artery disease
  5. coronary CT
  6. coronary CTA
  7. fractional flow reserve
  8. quantitative coronary angiography

Authors

Affiliations

Isabella Lipkin, BA
The George Washington University School of Medicine and Health Sciences, 2150 Pennsylvania Ave NW, Ste 4-417, Washington, DC 20037.
Anha Telluri, BA
The George Washington University School of Medicine and Health Sciences, 2150 Pennsylvania Ave NW, Ste 4-417, Washington, DC 20037.
Yumin Kim, BA
The George Washington University School of Medicine and Health Sciences, 2150 Pennsylvania Ave NW, Ste 4-417, Washington, DC 20037.
Alfateh Sidahmed, MD
The George Washington University School of Medicine and Health Sciences, 2150 Pennsylvania Ave NW, Ste 4-417, Washington, DC 20037.
Joseph M. Krepp, MD
The George Washington University School of Medicine and Health Sciences, 2150 Pennsylvania Ave NW, Ste 4-417, Washington, DC 20037.
Brian G. Choi, MD
The George Washington University School of Medicine and Health Sciences, 2150 Pennsylvania Ave NW, Ste 4-417, Washington, DC 20037.
Rebecca Jonas, MD
Jefferson Medical Institute, Philadelphia, PA.
Hugo Marques, MD, PhD
Faculdade de Medicina da Universidade Católica Portuguesa, Lisboa, Portugal.
Hyuk-Jae Chang, MD, PhD
Division of Cardiology, Severance Cardiovascular Hospital and Severance Biomedical Science Institute, Yonsei University College of Medicine, Yonsei University Health System, Seoul, South Korea.
Jung Hyun Choi, MD, PhD
Ontact Health, Inc., Seoul, South Korea.
Joon-Hyung Doh, MD
Division of Cardiology, Inje University Ilsan Paik Hospital, Goyang-si, South Korea.
Ae-Young Her, MD
Kang Won National University Hospital, Chuncheon, South Korea.
Bon-Kwon Koo, MD, PhD
Department of Internal Medicine, Seoul National University Hospital, Seoul, South Korea.
Chang-Wook Nam, MD, PhD
Cardiovascular Center, Keimyung University Dongsan Hospital, Daegu, South Korea.
Hyung-Bok Park, MD
Department of Internal Medicine, Division of Cardiology, International St. Mary's Hospital, Catholic Kwandong University College of Medicine, Incheon, South Korea.
Sang-Hoon Shin, MD
Department of Internal Medicine, Division of Cardiology, Ewha Women's University Seoul Hospital, Seoul, South Korea.
Jason Cole, MD
Mobile Cardiology Associates, Mobile, AL.
Alessia Gimelli, MD
Department of Imaging, Fondazione Toscana Gabriele Monasterio, Pisa, Italy.
Muhammad Akram Khan, MD
Cardiac Center of Texas, McKinney, TX.
Bin Lu, MD
State Key Laboratory of Cardiovascular Disease, Fuwai Hospital, Beijing, China.
Yang Gao, MD
Cardiovascular Center, St. Luke's International Hospital, Tokyo, Japan.
Faisal Nabi, MD
Houston Methodist DeBakey Heart and Vascular Center, Houston, TX.
Ryo Nakazato, MD, PhD
Cardiovascular Center, St. Luke's International Hospital, Tokyo, Japan.
U. Joseph Schoepf, MD
Medical University of South Carolina, Charleston, SC.
Roel S. Driessen, MD
Amsterdam University Medical Center, VU University Medical Center, Amsterdam, The Netherlands.
Michiel J. Bom, MD
Amsterdam University Medical Center, VU University Medical Center, Amsterdam, The Netherlands.
James J. Jang, MD
San Jose Medical Center, Kaiser Permanente Hospital, San Jose, CA.
Michael Ridner, MD
Heart Center Research, LLC, Huntsville, AL.
Chris Rowan, MD
Renown Heart and Vascular Institute, Reno, NV.
Erick Avelar, MD
Oconee Heart and Vascular Center, St. Mary's Hospital, Athens, GA.
Philippe Généreux, MD
Gagnon Cardiovascular Institute at Morristown Medical Center, Morristown, NJ.
Paul Knaapen, MD, PhD
Amsterdam University Medical Center, VU University Medical Center, Amsterdam, The Netherlands.
Guus A. de Waard, MD
Amsterdam University Medical Center, VU University Medical Center, Amsterdam, The Netherlands.
Gianluca Pontone, MD, PhD
Centro Cardiologico Monzino, IRCCS, Milan, Italy.
Daniele Andreini, MD, PhD
Centro Cardiologico Monzino, IRCCS, Milan, Italy.
Department of Biomedical and Clinical Sciences “Luigi Sacco,” University of Milan, Milan, Italy.
Mouaz H. Al-Mallah, MD
Houston Methodist DeBakey Heart and Vascular Center, Houston, TX.
Tami R. Crabtree, MS
Cleerly Inc., New York, NY.
James P. Earls, MD
The George Washington University School of Medicine and Health Sciences, 2150 Pennsylvania Ave NW, Ste 4-417, Washington, DC 20037.
Cleerly Inc., New York, NY.
Andrew D. Choi, MD [email protected]
The George Washington University School of Medicine and Health Sciences, 2150 Pennsylvania Ave NW, Ste 4-417, Washington, DC 20037.
James K. Min, MD
Cleerly Inc., New York, NY.

Notes

Address correspondence to A. D. Choi ([email protected], @AChoiHeart).
H. Marques is a paid consultant for and owns equity in Cleerly, Inc. T. R. Crabtree is an employee of Cleerly, Inc. J. P. Earls is an employee of and owns equity in Cleerly, Inc. A. D. Choi owns equity in Cleerly, Inc. J. K. Min is an employee of and owns equity in Cleerly, Inc.; serves on the scientific advisory board for Arineta; and owns equity in Upside Foods. The remaining authors declare that they have no disclosures relevant to the subject matter of this article.

Funding Information

Supported by grant funding from the GW Heart and Vascular Institute (to A. D. Choi) and research funding from NIH (to J. K. Min).

Metrics & Citations

Metrics

Citations

Export Citations

To download the citation to this article, select your reference manager software.

Articles citing this article

Media

Figures

Other

Tables

Share

Share

Copy the content Link

Share on social media