Enhancing the diagnosis of functionally relevant coronary artery disease with machine learning.

Functionally relevant coronary artery disease (fCAD) can result in premature death or nonfatal acute myocardial infarction. Its early detection is a fundamentally important task in medicine. Classical detection approaches suffer from limited diagnostic accuracy or expose patients to possibly harmful radiation. Here we show how machine learning (ML) can outperform cardiologists in predicting the presence of stress-induced fCAD in terms of area under the receiver operating characteristic (AUROC: 0.71 vs. 0.64, p = 4.0E-13). We present two ML approaches, the first using eight static clinical variables, whereas the second leverages electrocardiogram signals from exercise stress testing. At a target post-test probability for fCAD of <15%, ML facilitates a potential reduction of imaging procedures by 15-17% compared to the cardiologist's judgement. Predictive performance is validated on an internal temporal data split as well as externally. We also show that combining clinical judgement with conventional ML and deep learning using logistic regression results in a mean AUROC of 0.74.

Combining anatomical and biochemical markers in the detection and risk stratification of coronary artery disease.

AIMS: We aimed to test the hypothesis if combining coronary artery calcium score (Ca-score) as a quantitative anatomical marker of coronary atherosclerosis with high-sensitivity cardiac troponin as a quantitative biochemical marker of myocardial injury provided incremental value in the detection of functionally relevant coronary artery disease (fCAD) and risk stratification. METHODS AND RESULTS: Consecutive patients undergoing myocardial perfusion single-photon emission computed tomography (MPS) without prior CAD were enrolled. The diagnosis of fCAD was based on the presence of ischaemia on MPS and coronary angiography; fCAD was centrally adjudicated in the diagnostic and prognostic domain. Diagnostic accuracy was evaluated using the area under the receiver-operating characteristic curve (AUC). The composite of cardiovascular death and non-fatal acute myocardial infarction (AMI) within 730 days was the primary prognostic endpoint. Among 1715 patients eligible for the diagnostic analysis, 399 patients had fCAD. The combination of Ca-score and high-sensitivity cardiac troponin T (hs-cTnT) had good diagnostic accuracy for the diagnosis of fCAD (AUC 0.79, 95% confidence interval (CI) 0.77-0.81), but no incremental value compared with the Ca-score alone (AUC 0.79, 95% CI 0.77-0.81, P = 0.965). Similar results were observed using high-sensitivity cardiac troponin I (AUC 0.80, 95% CI 0.77-0.82) instead of hs-cTnT. Among 1709 patients (99.7%) with available follow-up, 59 patients (3.5%) suffered the composite primary prognostic endpoint (non-fatal AMI, n = 34; CV death, n = 28). Both Ca-score and hs-cTnT had independent prognostic value. Increased risk was restricted to patients with elevation in both markers. CONCLUSION: The combination of the Ca-score with hs-cTnT increases the prognostic accuracy for future events but does not provide incremental value vs. the Ca-score alone for the diagnosis of fCAD. STUDY REGISTRATION: Clinical trial registration: NCT00470587.