Study Population
This study was a single-centre retrospective study approved by the Institutional Review Board (IRB)/ Ethics committee (Central Office for Research Ethics Committee Reference 2020-P2-010-02). All patients signed informed consent forms and agreed to use of their data for further research. Between January 2017 and November 2018, 252 consecutive patients (aged over 18 years old) with suspected or known coronary heart disease (CHD) underwent both CCTA and ICA examinations within six months. Patients with iodine contrast agent allergy, atrial fibrillation, renal failure and pregnancy were excluded from this study. Of these initially included patients, seventeen CCTA or ICA data were incomplete, 6 had abnormal coronary origins or had undergone bypass surgery, 7 had poor image quality, and 26 had three vessels (left anterior descending (LAD) artery, left circumflex (LCx) artery and right coronary artery (RCA)) that could not be evaluated simultaneously; these patients were excluded. The reasons were as follows: severe extensive calcification[19], stents and motion artefacts. Finally, 196 patients were enrolled (Figure 1). ICA is the gold standard of diagnosis and interpreted jointly for stenosis ≥ 50% by an expert panel of three cardiovascular experts with at least 10 years of experience in both ICA and CCTA.
CCTA Image Acquisition
A 256-section CT (GE Healthcare, Waukesha, Wisconsin, US), a 64-section CT (GE Healthcare, Waukesha, Wisconsin, US) and a 128-row multidetector CT (Philips Medical Systems, Eindhoven, The Netherlands) were used to capture patient image data. Prospective electrocardiographic gating was employed. IoproMide (IoproMide, Ultravist 370; Bayer Healthcare LLC, Whippany, New Jersey) or Iohexol (Omnipaque 350, GE Healthcare, Princeton, NJ) was injected at 5-6 ml/s into the antecubital vein. All scanners had a layer thickness and spacing of 0.625 mm.
CCTA Analysis
All 196 CCTA patient datasets were reconstructed at a workstation (GE Advantage Workstation 4.6 or 4.7, GE Healthcare, Waukesha, Wisconsin), which could reconstruct the data into multiplanar reformation (MPR) and curved MPR (cMPR) images based on the original axial image, and then these images were transferred to a picture archiving and communication system (PACS). Patients were defined as positive when stenosis ≥ 50% as significant coronary artery disease. Six readers interpreted the stenosis via CCTA for all patients. Readers had different levels of experience: four readers (Reader 1 to Reader 4) were general radiologists with less experience in cardiovascular imaging and had interpreted less than 50 cases of coronary artery stenosis via CCTA; none of them had been mentored [20]. Reader 5 and Reader 6 were cardiovascular radiologists with at least 5 years of experience with CCTA, corresponding to level II competency (independent practitioners, IP) [21]. Inexperienced Reader 1 and Reader 2 and experienced Reader 5 and Reader 6 evaluated all patient data without the AI system on the same PACS. Inexperienced Reader 3 and Reader 4 evaluated the same patient data on the AI workstation and were assisted by AI in the diagnosis of coronary stenosis. In this study, we did not make the same readers interpret the 196 CCTA datasets with and without AI because CCTA interpretation experience is related to the number of cases [7, 20, 22], 196 CCTA cases had a significant impact on the reader’s results, therefore, we selected four readers with the similar experience rather than the same reader to interpret the data separately with and without aid by the AI system and the reader recall bias was effectively avoided. This AI system could independently perform automatic reconstruction and intelligently diagnose coronary artery stenosis. All readers analysed the four primary coronary arteries, the left main (LM) artery, LAD, LCx and RCA, and recorded the presence of vessels and whether ≥ 50% stenosis was present. Vessels with severely extensive calcified plaque, stents or a diameter of ≤ 1.5 mm were excluded from this study.
AI System
Data acquisition
The AI system used was “CoronaryDoc clinical decision Support Plat- form V1.0′′ from Shukun (Beijing) Technology Co., Ltd [23]. All CCTA data were transferred from a GE Advantage Workstation 4.6 or 4.7 to the AI workstation, and then the AI system extracted the centerline [24] and automatically reconstructed MPR and cMPR images based on the original axial image.
Coronary artery segmentation and naming
The coronary artery segments were divided into 18 segments according to the Society of Cardiovascular Computed Tomography (SCCT) criteria [25]. The coronary tree segmentation architecture used was an improved 3-dimensional (3D) U-Net. The AI system used an automatic identification algorithm to achieve segmentation and naming of the coronary arteries [24].
Automatic reconstruction and intelligent diagnosis of coronary artery stenosis
The system was based on coronary tree segmentation, MPR, straightened rendering (SR), cMPR, maximum intensity projection (MIP) and volume rendering (VR) images, and it automatically reconstructed the data. Stenosis along the long axis of the vessel was calculated based on the radius of the lumen where the plaque was located and the radius of the upstream and downstream blood vessels (details in the supplementary material Appendices).
Statistical Analysis
All data were analysed using SPSS software (version 26.0, IBM, Washington, USA) and MedCalc software (version 19.0.4, bvba, Ostend, Belgium). Assuming a 60% prevalence of coronary artery disease at the patient level in our single centre, the area under the receiver operating characteristic curve (AUC) of patients without AI and with AI was 68% and 77%, respectively, and the sample size was estimated to be 143 patients with 86 disease cases and 57 disease-free cases. Categorical variables are shown as percentages, and continuous variables are shown as means and ranges. We used sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and the AUC to describe diagnostic performance and accuracy. ICA was used as the gold standard to assess the results of 6 readers in the detection of coronary artery stenosis ≥ 50% at the patient and vessel levels. AUC comparisons for the 6 readers were performed using the method of DeLong et al. [26]. We used Cohen's kappa coefficient to evaluate interobserver consistency for detecting ≥ 50% stenosis lesions among the inexperienced readers without AI, the inexperienced readers aided by AI and the two cardiovascular radiologists. Vessels not evaluated by the six readers or the AI system were not statistically analysed in this study. A P-value < 0.05 was considered statistically significant.