Deep convolutional neural networks to predict cardiovascular risk from computed tomography.

Coronary artery calcium is an accurate predictor of cardiovascular events. While it is visible on all computed tomography (CT) scans of the chest, this information is not routinely quantified as it requires expertise, time, and specialized equipment. Here, we show a robust and time-efficient deep learning system to automatically quantify coronary calcium on routine cardiac-gated and non-gated CT. As we evaluate in 20,084 individuals from distinct asymptomatic (Framingham Heart Study, NLST) and stable and acute chest pain (PROMISE, ROMICAT-II) cohorts, the automated score is a strong predictor of cardiovascular events, independent of risk factors (multivariable-adjusted hazard ratios up to 4.3), shows high correlation with manual quantification, and robust test-retest reliability. Our results demonstrate the clinical value of a deep learning system for the automated prediction of cardiovascular events. Implementation into clinical practice would address the unmet need of automating proven imaging biomarkers to guide management and improve population health.

Incremental Prognostic Value of Myocardial Blood Flow Quantified With Stress Dynamic Computed Tomography Perfusion Imaging.

OBJECTIVES: This study aimed to evaluate whether myocardial blood flow (MBF) quantified with dynamic computed tomography perfusion imaging (CTP) has an incremental prognostic value over coronary CT angiography (CTA) for major adverse cardiac events (MACEs) in patients with suspected coronary artery disease (CAD). BACKGROUND: The incremental prognostic value of CTP over CTA is unclear. The quantification of MBF with dynamic CTP may potentially enhance risk stratification. METHODS: A total of 332 patients (67% men; age: 67 ± 10 years) with suspected CAD who underwent CTA and dynamic CTP was analyzed. A MACE was defined as cardiac death, nonfatal myocardial infarction (MI), unstable angina, or hospitalization for congestive heart failure. A summed stress score (SSS) was calculated by adding scores of all myocardial segments according to normalized MBF values. Abnormal perfusion was defined as SSS ≥4. Obstructive CAD was defined as ≥50% stenosis in ≥1 vessel on CTA. RESULTS: During a median follow-up of 2.5 years, 19 patients had a MACE. Multivariate analysis showed that, when adjusted for obstructive CAD on CTA, abnormal perfusion was significantly associated with hazards for MACEs (hazard ratio [HR]: 5.7; 95% confidence interval [CI]: 1.9 to 16.9; p = 0.002), with a significant improvement in the prognostic value. Abnormal perfusion was an independent predictor even when adjusted for ≥70% stenosis in ≥1 vessel (HR: 5.4; 95% CI: 1.7 to 16.7; p = 0.003) or adjusted for ≥50% stenosis in ≥2 vessels (HR: 6.5; 95% CI: 2.2 to 18.9; p = 0.001). In the setting of obstructive CAD, annualized event rates showed a significant difference between the patients with and without abnormal perfusion for all events (12.2% vs. 1.5%; p = 0.002) and for cardiac death and nonfatal MI (4.2% vs. 0%; p = 0.015). CONCLUSIONS: MBF quantified with dynamic CTP has an incremental prognostic value over CTA. The addition of dynamic CTP to CTA allows improved risk stratification of patients with CTA-detected stenosis.

Diagnostic Accuracy of Endocardial-to-Epicardial Myocardial Blood Flow Ratio for the Detection of Significant Coronary Artery Disease With Dynamic Myocardial Perfusion Dual-Source Computed Tomography.

BACKGROUND: Previous dynamic stress computed tomography perfusion (CTP) studies used absolute myocardial blood flow (MBF in mL/100 g/min) as a threshold to discriminate flow-limiting coronary artery disease (CAD), but absolute MBF can be vary because of multiple factors. The aim of this study was to compare the diagnostic performance of absolute MBF and the transmural perfusion ratio (TPR) for the detection of flow-limiting CAD, and to clarify the influence of CT delayed enhancement (CTDE) on the diagnostic performance of CTP.Methods and Results:We retrospectively enrolled 51 patients who underwent dual-source CTP and invasive coronary angiography (ICA). TPR was defined as the endocardial MBF of a specific segment divided by the mean of the epicardial MBF of all segments. Flow-limiting CAD was defined as luminal diameter stenosis >90% on ICA or a lesion with fractional flow reserve ≤0.8. Segmental presence and absence of myocardial scar was determined by CTDE. The area under the receiver-operating characteristics curve (AUC) of TPR was significantly greater than that of MBF for the detection of flow-limiting CAD (0.833 vs. 0.711, P=0.0273). Myocardial DE was present in 27 of the 51 patients and in 34 of 143 territories. When only territories containing DE were considered, the AUC of TPR decreased to 0.733. CONCLUSIONS: TPR calculated from absolute MBF demonstrated higher diagnostic performance for the discrimination of flow-limiting CAD when compared with absolute MBF itself.

Comparison of Displacement Encoding With Stimulated Echoes to Magnetic Resonance Feature Tracking for the Assessment of Myocardial Strain in Patients With Acute Myocardial Infarction.

The aim of this study was to compare myocardial strain by cardiovascular magnetic resonance feature tracking (CMR-FT) to those derived from displacement encoding with stimulated echoes (DENSE) in patients with acute myocardial infarction (AMI). Twenty patients (65 pa13 years) with AMI underwent cine, DENSE, black-blood T2-weighted and late gadolinium enhancement CMR at 1.5 T. Global and segmental strain was determined by CMR-FT analysis and DENSE on matched 3 short-axis planes. Global circumferential strain by CMR-FT showed a good agreement with that by DENSE (r = 0.85, p <0.001; bias 0.02, limits of agreement -0.03 to 0.06). For segmental circumferential strain, r coefficient between CMR-FT and DENSE was 0.61 (p <0.001) with bias of 0.02, limits of agreement of -0.07 to 0.11. Regional circumferential strain determined by CMR-FT in infarct segments (-0.08 ± 0.05) was significantly altered compared with that in remote normal segments (-0.15 ± 0.05, p <0.001). CMR-FT measurement of regional and global circumferential strain showed good agreement with DENSE in patients with AMI.

Underestimation of myocardial blood flow by dynamic perfusion CT: Explanations by two-compartment model analysis and limited temporal sampling of dynamic CT.

PURPOSE: Previous studies using dynamic perfusion CT and volume perfusion CT (VPCT) software consistently underestimated the stress myocardial blood flow (MBF) in normal myocardium to be 1.1-1.4 ml/min/g, whilst the O 15-water PET studies demonstrated the normal stress MBF of 3-5 ml/min/g. We hypothesized that the MBF determined by VPCT (MBF-VPCT) is actually presenting the blood-to-myocardium transfer constant, K1. In this study, we determined K1 using Patlak plot (K1-Patlak) and compared the results with MBF-VPCT. MATERIAL AND METHODS: 17 patients (66 ± 9 years, 7 males) with suspected coronary artery disease (CAD) underwent stress dynamic perfusion CT, followed by rest coronary CT angiography (CTA). Arterial input and myocardial output curves were analyzed with Patlak plot to quantify myocardial K1. Significant CAD was defined as >50% stenosis on CTA. A simulation study was also performed to investigate the influence of limited temporal sampling in dynamic CT acquisition on K1 using the undersampling data generated from MRI. RESULTS: There were 3 patients with normal CTA, 7 patients with non-significant CAD, and 7 patients with significant CAD. K1-patlak was 0.98 ± 0.35 (range 0.22-1.67) ml/min/g, whereas MBF-VPCT was 0.83 ± 0.23 (range 0.34-1.40) ml/min/g. There was a linear relationship between them: (MBF-VPCT) = 0.58 x (K1-patlak) + 0.27 (r(2) = 0.65, p < 0.001). The simulation study done on MRI data demonstrated that Patlak plot substantially underestimated true K1 by 41% when true K1 was 2.0 ml/min/g with the temporal sampling of 2RR for arterial input and 4RR for myocardial output functions. CONCLUSIONS: The results of our study are generating hypothesis that MBF-VPCT is likely to be calculating K1-patlak equivalent, not MBF. In addition, these values may be substantially underestimated because of limited temporal sampling rate.

An elderly patient with severe aortic stenosis and myocardial infarction with a huge mobile thrombus as complication in the left ventricle.

An 86-year-old woman was admitted for emergency treatment of increasing dyspnea. Transthoracic echocardiography revealed decreased left ventricular systolic function with dyskinesis at the apex, and severe aortic stenosis. The apex of the left ventricle showed a huge mobile thrombus. Coronary angiography revealed total occlusion at the middle portion of the left anterior descending coronary artery. Emergency operation was successful, and a partially calcified thrombus was observed at the site of the old myocardial infarction area. In this case, myocardial infarction and elevated intraventricular pressure due to aortic stenosis likely contributed to the wall motion abnormality and thrombus formation.