This single-center, retrospective study was approved by the ethics committee of our hospital, and the requirement for written informed consent was waived owing to the retrospective nature of the study. Patients with suspected coronary heart disease due to chest pain who underwent CCTA at our center between December 2019 and September 2021 were consecutively selected from the hospital radiology information system. Inclusion criteria included CCTA imaging performed using the SSF technique with CT (GE Healthcare). Exclusion criteria included the image that does not meet the diagnostic requirements due to motion artifacts. A total of 90 patients (23 women; mean age: 67.1 ± 11.7 [range 38–90] years) were enrolled for image quality evaluation and 35 of the 90 patients who undergone digital subtraction angiography (DSA) were also used for diagnostic accuracy evaluation (Figure 1). Of these 35 patients, 28 patients' main lesion caused significant coronary stenosis is mixed plaque and 7 patients' is non-calcified plaque.

All patients underwent wide-coverage-detector CT (Revolution CT, GE Healthcare) with the following parameters: tube voltage, 100–120 kV; automatic tube current modulation, 300–800 mA; data acquisition window (R–R interval), 30–80%; and detector coverage, 100 mm, 120 mm, 140 mm, or 160 mm dependent on patient heart sizes. The gantry rotation speed was 0.28 s/rotation.

CCTA images with a slice thickness of 0.625 mm and increment of 0.625 mm were reconstructed using the 80% adaptive statistical iterative reconstruction-V [ASIR-V80%, GE Healthcare] with the STND kernel at the optimal reconstruction phase automatically selected by the SmartPhase technique (GE Healthcare), and the cardiac motion correction technique (snapshot freeze, SSF) was also applied (12, 15, 17). The generated images were assigned as the routine group. Subsequently, all the CCTA images of the routine group were sent to an Advantage Workstation (AW4.7, GE Healthcare) to undergo further image filtering. After our preliminary exploration and comparison with different combinations of smooth and edge enhancement filters that are readily available on the AW4.7, the filter combination was narrowed down to the combination of smooth 3D+ and edge-2. On the workstation, the volume rendering software was selected, and the Smooth 3D+ image filter option was selected for noise reduction. The resulting images were subsequently saved as a new group named Smooth 3D+. In a second pass, volume rendering software was once again selected, and the Edge-2 image filter option was chosen for edge sharpening of the Smooth 3D+ group. The resulting images were saved as a new group named the smooth 3D+ and edge-2 group (SE group). The signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and image quality scores were then compared between the original routine group and the SE group (Figure 2).

Quantitative measurements were performed including CT values and standard deviation (SD) of the lumen of the aortic root (AO), proximal left anterior descending artery (LAD-P), middle left anterior descending artery (LAD-M), distal left anterior descending artery (LAD-D), proximal left circumflex artery (LCX-P), middle left circumflex artery (LCX-M), distal left circumflex artery (LCX-D), proximal right coronary artery (RCA-P), middle right coronary artery (RCA-M), distal-proximal right coronary artery (RCA-D), and perivascular adipose tissue (PVAT). The region of interest (ROI) in the AO was set to 90 mm², and those in others were set to 1 mm². Image noise was classified as the SD of the attenuation within the ROI in the AO. Each ROI was measured three times by an independent reader, and the average of the three measurements was used for further analysis. The SNR was defined as CT value (vessel)/image noise, and the CNR was defined as [CT value (vessel) - CT value (PVAT)]/SD(PVAT).

Double-blinded image quality scoring was performed by two radiologists, one with 8 years of experience and the other with more than 15 years of experience in cardiovascular diagnosis, using the 18-segment model of the Society of Cardiovascular Computed Tomography guidelines (25). Evaluations were conducted over 1 month to eliminate the observers' recall effects between routine and SE groups of the same patient. A 4-point scoring system was used for overall image quality analysis (26, 27) [(1) excellent, no artifacts; (2) good, minor artifacts; (3) moderate, artifacts, but diagnosis still possible; and 4: poor, non-diagnostic]. Each patient's CCTA was evaluated through the following 10 segments for the severity of stenosis: left main artery, LAD-P, LAD-M, LAD-D, LCX-P, LCX-D, RCA-P, RCA-M, RCA-D, and posterior descending artery. The stenosis degree was assessed using Coronary Artery Disease - Reporting and Data System (CAD-RADS) with a 7-grade scoring system (0–2: non-significant stenosis [luminal irregularities or lumen diameter narrowing <50%]; 3–5: significant stenosis [lumen diameter narrowing ≥50%]; 6: non-diagnostic situation) (Table 1). In cases of disagreement on the image quality score or stenosis assessment between the two radiologists, a third experienced observer (more than 15 years of experience) would mediate to reach a final agreement. Individual adjustments of the window center and window width were allowed.

The results of digital subtraction angiography (DSA) were used as the reference standard to determine the significant stenosis independent of the CCTA results. Patients underwent DSA for diagnostic accuracy by two cardiologists (with more than 10 years of experience). The stenosis degree of each segment of coronary vessels (range: 0–100%) was quantitatively evaluated. Similar to CCTA, all segments were classified as being non-significant stenosis, significant stenosis, or non-diagnostic situation.

An image quality testing software (Imatest, V5.2.21, USA) was used to evaluate the spatial frequency response reflecting the image sharpness through modulation transfer function (MTF) in the 2D image domain (28). One radiologist chose the same Digital Imaging and Communications in Medicine (DICOM) slices of the beginning of the coronary sinus of every patient in each group (one from the routine group and one from the SE group) and then exported this image as a 2D image (Portable Network Graphic, PNG) for further image sharpness evaluation. Using the Imatest software, the same ROI was set on the same position of two 2D images, including the heart, and the MTF50, MTF50P, and MTF30 were then calculated automatically by the software (Figure 3).

The radiation dose of CCTA imaging, including the dose-length product (DLP) and the volume CT dose index was recorded through the dose reports provided by the CT scanner. The effective dose was estimated by multiplying the DLP by a conversion factor of 0.014 mSv/(mGy·cm) (29).

SPSS Statistics V22.0 (IBM, Chicago, IL, USA) was used for statistical analysis. Quantitative indices are presented as mean ± SD and medians (range: maximum to minimum). Data with non-normal distribution are expressed as medians (interquartile ranges). Quantitative data were analyzed using the independent samples t-test or the Wilcoxon signed rank-sum test. Counts were assessed using Pearson's chi-squared test, and Fisher's exact test was employed to examine the probability when the expected value was <5. A kappa value was calculated and defined as follows: < 0.20, almost inconsistent; 0.21–0.40, slightly consistent; 0.41–0.60, medium consistency; 0.61–0.80, good consistency; and 0.81–1.00, perfect consistency (27). The false-positive rate, false-negative rate, positive predictive value, negative predictive value (NPV), sensitivity, specificity, and accuracy were calculated for CCTA vs. DSA for each reconstruction on per-vessel and total levels and analyzed by the χ² test, and differences were analyzed with ANOVA between three groups. The significance level was set at p < 0.05.

The CT value of the AO ROI and the image noise in the routine group were higher than those in the SE group (P < 0.001). Moreover, the SNR and CNR were significantly higher in the SE group than in the routine group (all Ps < 0.05). The SE group had significantly higher MTF50 and MTF50P values than the routine group, while the MTF30 values were similar in these two groups (Table 3).

Due to individual differences in coronary artery anomalies, the image quality score was calculated based on three main coronary arteries (i.e., LAD, LCX, and RCA). The consistency of the two readers was good or perfect for each vessel (LAD, LCX, and RCA) and total vessels (kappa: 0.71–0.84), and the image quality scores in the SE group were significantly lower (i.e., better) than those in the routine group in each vessel and total vessels (all Ps < 0.001). Image quality scores of three main coronary arteries (LAD, LCX, and RCA) and total segments improved to different degrees after removing image quality score 1 of those segments (percent: 49–70%) (Table 4).

Among the 90 patients, 35 underwent DSA for stenosis evaluations after CCTA imaging. Based on the results of DSA, the accuracy of the findings related to stenosis in these 35 patients was compared among the three groups (routine group, SE group, and DSA group). The results showed no statistically significant difference in the accuracy of the stenosis evaluations between the SE and routine groups at the per-vessel and total levels (all Ps > 0.05). However, the SE group showed a slightly higher NPV on the LAD and RCA stenosis evaluations, as well as total NPV. The SE group also showed slightly higher sensitivity and accuracy than the routine group on the RCA stenosis evaluation. Moreover, eight segments were evaluated with CAD-RADS 6 (non-diagnostic situation) in the routine group, while two segments of RCA from these eight segments were improved to the diagnostic situation (Table 5).