2.1 Patients
In the present study, 123 patients with known or suspected CAD who underwent CCTA and subsequent ICA between July 2021 and June 2022 were screened. The exclusion criteria were as follows: patients with a history of coronary stenting or coronary artery bypass grafting, renal insufficiency with a baseline creatinine level of > 2.0 mg/dL, congestive heart failure, and cardiogenic shock; patients with suboptimal image quality were also excluded. A total of 94 patients with a diagnosis of known or suspected CAD as per the registry were included in the final analysis. The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013), and was approved by the Medical Ethics Committee. The need for informed consent was waived for this retrospective study. Figure 1 shows a detailed time line of the study protocol.
2.2 Coronary Cta
The Toshiba 320-row CT (Aquilion ONE, Toshiba, Tokyo, Japan) scanner was used for data acquisition. In patients with heart rates exceeding 80 beats/min, 25–50 mg metoprolol was administrated orally; scanning was performed after the heart rate was reduced to below 70 beats/min. An indwelling 18-G trocar was placed in the antecubital vein, and non-ionic iodine contrast medium (350 mg/L) was injected using a two chamber high pressure injector (Mallinckrodt Pharmaceuticals, Staines-upon-Thames, UK); patients with a body mass index of < 25 kg/m2 were administered 60 mL of the contrast medium at a rate of 4.0-4.5 mL/s, while those with a body mass index of 25–35 kg/m2 were administered a volume of 70 mL at a rate of 4.5-5.0 mL/s. The 320 detector-row CT images were acquired with a gantry rotation time of 350 ms and a tube voltage of 120 kV; the radiation dose was 1.0 ± 0.5 mSv. The best phase was reconstructed retrospectively, with reformat thicknesses and intervals of 0.5 mm, each. The collected data were transferred to the Vitrea FX workstation (Canon Medical, Japan) for post-processing; this included multi-planar reformatting, curved multi-planar reformatting, and volume reformatting, among others.
Image quality was categorized into four grades: poor, medium, good, and excellent. Only images with good to excellent quality were included for further analysis. All CT images were analyzed separately by two experienced senior radiologists. In cases where there was a difference in opinion, a final conclusion was drawn after detailed discussion. The coronary tree was assessed based on the 18-segment coronary artery segmentation system published by the American Society for Cardiovascular Computed Tomography in 2014; the severity of stenosis at each segment was assessed visually, and patients were classified according to the Coronary Artery Disease Reporting and Data System [16]. Stenosis degree was categorized into six grades, as follows: 0, no stenosis; 1, slight stenosis (stenosis < 25%); 2, mild stenosis (stenosis degree 25% ~ 49%); 3, moderate stenosis (stenosis degree 50%-69%); 4. Severe stenosis (70% ~ 99% stenosis); and 5, occlusion (stenosis degree 100%) [5]. Obvious obstructive CAD was defined by stenosis of ≥ 50% of the vessel diameter.
2.3 Ct-ffr Analysis
Artificial intelligence deep learning-based software (Shukun Technology Co., Ltd.) was utilized for the calculation of CT-FFR in this study[11]. This deep learning algorithm calculates the CT-FFR from the reduced-order method combined with machine learning modification[9; 18]. The centerline is completely automatic and no manual related to changing something is needed though the whole pipeline. No user action that helps a bad situation also means that the results are working regularly without black-and-white bias.
All CCTA data were transmitted to the cloud workstation in standard Digital Imaging and Communications in Medicine format. The software calculated and obtained the blood flow reserve fraction based on the training and learning of fluid dynamics simulation data from a large number of coronary cases; the output time was < 10 min per case. The software provided a specific three-dimensional model of the coronary arteries, allowing the investigator to obtain CT-FFR values along any given point along the length of the coronary vessels.
The lesion-specific CT-FFR (at 2 cm distal to the stenosis) after coronary artery stenosis was recorded for each vessel; if the CT-FFR value was ≤ 0.8, the patient was classified as having obstructive CAD (Fig. 2). The distal-tip CT-FFR (the CT-FFR value at the distal end of each vessel) was also recorded using the same ischemia threshold corresponding to the location of coronary artery stenosis and corresponding 3D coronary artery model (Fig. 2).
2.4 Fai Analysis
FAI analysis closely followed the methodology previously defined by Oikanomou et al.; it was performed using CoronaryDoc (Shukun Technology). The measurement of PCAT attenuation around the proximal right coronary artery (RCA) is a standardized method; this parameter has been used in prior studies as a representative biomarker of coronary inflammation [19]. The proximal 40-mm segment of the left anterior descending coronary artery (LAD), left circumflex coronary artery (LCx), and RCA were traced, as previously described [2; 10]. The lumen and inner and outer vessel wall borders were tracked within the pre-identified segment of interest in an automated manner with additional manual optimization. PCAT was defined as the adipose tissue located around the outer vessel wall, within a radial distance equal to the coronary vessel diameter. Voxel histograms of CT attenuation were plotted and the mean CT attenuation of all voxels ranging between − 190 to − 30 HU (threshold used for defining adipose tissue [19]) within the PCAT volume was calculated. The FAI was defined as the mean CT attenuation of PCAT in the traced 40-mm segment on crude analysis. Representative images of FAI analysis are shown in Fig. 2. PCAT analysis was performed by investigators who were blinded to clinical data.
2.5 Ica And Revascularization
The primary endpoint in our study was revascularization during or directly after referral for ICA; this included percutaneous coronary revascularization and coronary artery bypass grafting. Secondary analysis involved the prediction of lesion-specific ischemia (FFR ≤ 0.80) and revascularization within the sub-cohort of patients who received invasive FFR during ICA. The procedure was performed by two cardiologists with at least 10 years’ experience.
Statistical Analysis
SPSS 20.0 (SPSS, Inc., Chicago, IL, USA) and MedCalc 15.6 (MedCalc Software, Mariakerke, Belgium) software packages were used for all statistical analyses. Normally and non-normally distributed continuous variables have been presented as means with standard deviation and medians with interquartile ranges, respectively. The Student’s t- and Mann-Whitney U tests were used for normally and non-normally distributed data, respectively; the chi-square test was used for categorical variables. Univariable and multivariable analyses were performed on both per-patient and per-vessel basis. Receiver operating characteristic (ROC) curves were constructed using the CT-FFR and FAI to establish the optimal threshold values for predicting revascularization using CCTA. The diagnostic capability was determined by calculating the area under the ROC curve (AUC); the DeLong method was used for comparison of AUCs. The optimal threshold value was determined according to the highest Youden’s J statistic. In cases where the optimal threshold value was adopted, the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy with corresponding 95% confidence intervals (CIs) were calculated for these quantitative parameters. P values < 0.05 were considered statistically significant.