Evaluation of a deep learning model on coronary CT angiography for automatic stenosis detection

PurposeThe purpose of this study was to evaluate a deep-learning model (DLM) for classifying coronary arteries on coronary computed tomography -angiography (CCTA) using the Coronary Artery Disease-Reporting and Data System (CAD-RADS).Materials and methodsThe DLM was trained with 10,800 curved multiplanar reformatted (cMPR) CCTA images classified by an expert radiologist using the CAD-RADS. For each of the three main coronary arteries, nine cMPR images 40° apart acquired around each arterial circumference were then classified by the DLM using the highest probability. For the validation set composed of 159 arteries from 53 consecutive patients, the images were read by two senior and two junior readers; consensus of the two seniors was the reference standard. With the DLM, the majority vote for the nine images was used to classify each artery. Three groups (CAD-RADS 0, 1–2, or 3–4–5) and 2 groups CAD-RADS 0–1–2 or 3–4–5 (<50% vs. ≥50% stenosis) were used for comparisons with readers and consensus. Performance of the model and readers was compared to the consensus reading using the intraclass coefficient (ICC) and Cohen's kappa coefficient at the artery and patient levels.ResultsWith the three groups at the artery level, the ICC of the DLM was 0.82 (95% CI: 0.75–0.88) and not significantly different from those of 3/4 readers; accuracy was 81%. With the binary classification, Cohen kappa coefficient of the DLM was 0.85 (95% CI: 0.69–0.94) and not significantly different from that of 3/4 readers; accuracy was 96%. At the patient level, sensitivity and specificity were 93% and 97% respectively, and the negative predictive value was 97%.ConclusionThe DLM detected ≥50% stenoses with performances similar to those achieved by senior radiologists.     

Coronary CTA enhanced with CTA based FFR analysis provides higher diagnostic value than invasive coronary angiography in patients with intermediate coronary stenosis

BackgroundCTA based FFR, a software based application, enhances diagnostic value of coronary computed tomography angiography (CTA) examination. However it remains unknown whether it improves accuracy over the gold standard of invasive coronary angiography (ICA) in predicting functionally significant coronary stenosis. The aim of our study was to compare diagnostic accuracies of coronary CTA, CTA based FFR, and ICA, with invasive FFR as the reference standard in patients with intermediate stenosis on CTA.Methods96 intermediate stenoses (50–90%) from 90 subjects, with intermediate pre-test probability of CAD, who underwent coronary CTA were analyzed. Each patient had subsequent ICA with FFR. CTA based FFR (cFFR v2.1, Siemens) analysis was performed on-site. The stenoses with invasive FFR≤0.8 were considered hemodynamically significant.Results41/96 stenoses were hemodynamically significant (FFR≤0.8). While the area under ROC curves (AUC) for identification of significant stenosis evaluated on QCA (0.653), visual ICA (0.652), qCTA (0.690) and visual CTA (0.660) did not significantly differ, the AUC for CTA based FFR (0.835) was significantly higher (p = 0.004, p = 0.004, p = 0.010, p = 0.007, respectively). The accuracies of CTA based FFR, qCTA and QCA were 76%, 63% and 58% respectively.ConclusionOur results suggest that diagnostic potential of routine coronary CTA, augmented with CTA based FFR analysis, is superior to ICA in patients with intermediate stenosis.     

Regadenoson dynamic computed tomography myocardial perfusion using low-dose protocol for evaluation of the ischemic burden. ULYSSES study

BackgroundThe purpose of this study was to assess the feasibility of low-dose dynamic regadenoson computed tomography perfusion (CTP) protocol, and to determine which parameters provide the best diagnostic yield for the presence and burden of ischemia in reference to the magnetic resonance myocardial perfusion imaging (MR MPI).MethodsFifty six patients with ≥1 intermediate (50–90%) coronary artery stenosis on CTA underwent dynamic stress CTP and MR MPI. The distribution of contrast agent in CTP was represented for each myocardial segment as either absolute or indexed: myocardial blood flow (MBF), myocardial blood volume (MBV), perfused capillary blood volume (PCBV), peak value (PV), time to peak (TTP), respectively.ResultsOf 56 patients (25 females, 63.5 ± 8.5y), 15 (27%) were diagnosed with reversible ischemia and 3 (5%) with fixed ischemia on the MR MPI. The median radiation dose for dynamic CTP scan was 352.00 [276.4–496.6] mGy*cm. The optimal cut-off point for the prediction of reversible ischemia on MR MPI for the absolute parameters were: MBF ≤156.49 (AUC=0.899), MBV ≤15.06 (AUC=0.901), PCBV ≤7.90 (AUC=0.880), PV ≤ 88.30 (AUC=0.766), TTP ≥22.58 (AUC=0.595); and for the indexed: indexed MBF ≤0.78 (AUC=0.926), indexed MBV ≤0.81 (AUC=0.924), indexed PCBV ≤0.70 (AUC=0.894); indexed PV ≤ 0.79 (AUC=0.869), indexed TTP ≤0.87 (AUC=0.685). The best parameters for ischemia detection were indexed MBF and indexed MBV, with sensitivities 91% and 89%, specificities 97% and 96%, NPV 99% and 99%, PPV 76% and 69%, and accuracies 96% and 95%, respectively. In per patient analysis, indexed MBF correlated significantly better with the ischemia burden than any of the absolute parameters (p < 0.01 for all comparisons).ConclusionsRegadenoson dynamic CTP using low-dose protocol is feasible while maintaining high diagnostic accuracy. The best diagnostic value may be provided by indexed parameters, of which indexed MBF and indexed MBV may provide best incremental value in identification of the presence and burden of ischemia.