Deep learning reconstruction allows for usage of contrast agent of lower concentration for coronary CTA than filtered back projection and hybrid iterative reconstruction

Abstract

Background

The demand for homogeneous and higher vascular contrast enhancement is critical to provide an appropriate interpretation of abnormal vascular findings in coronary computed tomography angiography (CTA).

Purpose

To evaluate the effect of various contrast media concentrations (Iohexol-370, Iohexol-300, Iohexol-240) and image reconstructions (filtered back projection [FBP], hybrid iterative reconstruction [IR], and deep learning reconstruction [DLR]) on coronary CTA.

Material and Methods

A total of 63 patients referred for coronary CTA between July and October 2021 were enrolled in this prospective study, and they randomly received one of three contrast media. CTA images were reconstructed with FBP, hybrid IR, and DLR. The CT attenuation, image noise, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) were calculated for all three images. The images were subjectively evaluated by two radiologists in terms of overall image quality, artifacts, image noise, and vessel wall delineation on a 5-point Likert scale.

Results

The application of DLR resulted in significantly lower image noise; higher CT attenuation, SNR, and CNR; and better subjective analysis among the three different concentrations of contrast media groups (P < 0.001). There was no significant difference in the CT attenuation of the left ventricle (P = 0.089) and coronary arteries (P = 0.072) between hybrid IR at Iohexol-300 and DLR at Iohexol-240. Furthermore, application of DLR to the Iohexol-240 significantly improved SNR and CNR; it achieved higher subjective scores compared with hybrid IR at Iohexol-300 (P < 0.001).

Conclusion

We suggest that using DLR with Iohexol-240 contrast media is preferable to hybrid IR with Iohexol-300 contrast media in coronary CTA.

Keywords

Contrast media image quality multidetector computed tomography image reconstruction coronary

Introduction

Computed tomography angiography (CTA) has been widely used owing to technological improvements in its ability to detect obstructive coronary artery disease. The demand for homogeneous and high-vascular contrast enhancement is critical for CTA to provide an appropriate interpretation of abnormal vascular findings (1,2). Coronary artery enhancement at a CT attenuation above 250 Hounsfield units (HU) is suggested for evaluating atherosclerotic changes and stenosis in coronary CTA (3). Generally, vascular enhancement is determined by scanning acquisition and reconstruction parameters (scan duration, tube voltage, and tube current), patient-related factors (heart rate, body weight, and cardiac output), injection rate, and contrast media concentration (4 –6). Although reducing the tube voltage can reduce the radiation dose and improve vascular attenuation, higher image noise and artifacts are major problems (7 –9). The linear relationship between the injection parameters of contrast media (total amount of contrast media, flow rate and concentration of contrast media, and saline chaser) and the degree of vascular attenuation has also been investigated in several studies (10,11). Increasing the flow rate of contrast media may result in contrast extravasation (12). Thus, a contrast media with higher iodine concentration is preferred to obtain a greater vascular enhancement. However, higher iodine concentration is not recommended for patients with cardiovascular disease because they are susceptible to contrast-induced nephropathy (13). Thus, an effective contrast media delivery method using a lower iodine concentration is preferable for the safety of patients and evaluation of abnormal changes in CTA.

The image reconstruction algorithm can improve the vascular enhancement without increasing the iodine concentration of the contrast media in CTA. Previous studies focused on the advantages of deep learning reconstruction (DLR). DLR was developed to reduce the image noise and improve spatial resolution through deep neural networks and its high-quality model-based iterative reconstruction (MBIR) or filtered back projection (FBP) training images, referred to as advanced intelligent clear-IQ engine (AiCE) (14) and TrueFidelity (15), respectively. Several clinical studies have demonstrated that the lesion detection rate and image quality of DLR are higher than those of iterative reconstruction (IR) for low-dose chest CT, abdominal CT, brain CTA, and coronary CTA (16 –21); however, there has been no evidence proving the effect of different contrast media iodine concentrations on DLR in coronary CTA. Enhancing vascular attenuation through DLR decreases the iodine load in patients with renal impairment.

The aim of the present study was to evaluate the effect of various contrast media concentrations (Iohexol-370, Iohexol-300, and Iohexol-240) on coronary CTA using the same flow rate and image reconstruction algorithms (FBP, hybrid IR, and DLR). Furthermore, this study identified the image reconstruction method that yields the highest image quality with the lowest possible iodine concentration.

Material and Methods

Participants

All individuals gave written informed consent for inclusion before they participated in this prospective study. The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of Inje University Seoul Paik Hospital (file no. 2021-06-009). A total of 63 patients (43 men; mean age = 59.32 years; age range = 26–86 years) referred for coronary CTA were enrolled in this study from July to October 2021 and randomly received one of three different contrast media (each was received by 21 patients). We used three different iodine concentrations of contrast media: Iohexol-370 (Bonorex 370; Central Medical Service, Republic of Korea); Iohexol-300 (Bonorex 300; Central Medical Service, Republic of Korea); and Iohexol-240 (Bonorex 240; Central Medical Service, Republic of Korea). The volume and rate of contrast media were consistent among the three different iodine concentrations. A total of 80 mL contrast media was injected into the antecubital vein at a rate of 4.5 mL/s via a power injector (CT Motion; Ulrich Medical, Ulm, Germany), followed by a 30-mL saline flush at a rate of 5.0 mL/s. The scan was triggered using the automatic bolus-tracking program (Surestart; Canon Medical Systems Corporation, Otawara, Japan) in the descending thoracic aorta (trigger threshold was set as 250 HU). Patients with a prior allergic reaction to iodine-containing compounds, pregnancy, impaired renal function (glomerular filtration rate <60 mL/min), hyperthyroidism, psychiatric disorder, or clinical instability were excluded.

Scanning protocols

All coronary CTAs were performed using a 320 multidetector volume scanner (Aquilion ONE Prism; Canon Medical Systems Corp., Otawara, Japan). Each patient underwent a non-contrast acquisition process for the analysis of coronary artery calcification, followed by coronary CTA. The following parameters were used for coronary CTA: detector collimation = 130 × 0.5 mm; field of view = 416 mm; gantry rotation time = 0.4 s; wide-volume scanning pitch = 0.813; slice thickness = 1 mm; and tube voltage = 100 kVP. Each section was 140 mm in the longitudinal direction. An automatic exposure control function (SD 20) was performed for tube current. All examinations were performed with prospective electrocardiography triggering methods covering 30%–85% of the cardiac cycle of the patients. The patients received no additional beta-blockers.

Image reconstruction

The raw data were retrieved from a CT scanner and reconstructed using the following three algorithms: FBP; hybrid IR with standard FC14 reconstruction kernel (adaptive iterative dose reduction 3-D [AIDR-3D] (17,22), Canon Medical Systems Corporation, Otawara, Japan); and DLR with body sharp option (AiCE; Canon Medical Systems Corporation, Otawara, Japan).

Objective analysis

The CT attenuation, image noise, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) were calculated at the same position in images reconstructed with FBP, hybrid IR, and DLR for each patient. The CT attenuation was measured by placing the region of interest (ROI) of 100 mm² in the left ventricle and ROI of 2 mm² in the proximal segments of the main coronary arteries, including the right coronary artery (RCA), left main coronary artery (LCA), left circumflex artery (LCX), and left anterior descending artery (LAD). The ROIs were placed in the center of the main coronary arteries while avoiding the inclusion of the vessel walls. The image noise determined from subcutaneous fat (ROI of 30 mm²) and muscle (ROI of 20 mm²) was used as a reference to calculate SNR and CNR. The SNR was calculated by dividing the attenuation of the main coronary arteries and left ventricle by the image noise. The CNR was assessed as the difference between the CT attenuation of the main coronary arteries and left ventricle and the attenuation of the muscle divided by the image noise. Histogram analysis was performed for the left ventricle, subcutaneous fat, and muscle by placing the ROI (100 mm², 30 mm², and 20 mm², respectively) in the adjacent seven slices; the mean and standard deviation were measured in HU. The histogram analysis was conducted offline using MATLAB version 8.2.0 (MathWorks, Natick, MA, USA).

Subjective analysis

The images were subjectively evaluated by two radiologists (with three and eight years of reading experience in diagnostic radiology). The observers were blinded to the image reconstruction methods and randomly evaluated the images in terms of overall quality, artifacts, and image noise. A 5-point Likert scale was used for the analysis (Table 1). The coronary arteries were divided into 15 segments according to the American Heart Association classification, and the vessel delineation was rated as follows: 5 = perfect vessel wall definition; 4 = very good vessel wall definition; 3 = good vessel wall definition; 2 = poor vessel wall definition; and 1 = insufficient vessel wall definition (23 –25).

Table 1.

Subjective image analysis criteria.

Score	Overall image quality	Artifacts	Image noise
5	Excellent	Absent	Minimal
4	Acceptable	Minor	Below average
3	Moderate	Moderate	Moderate
2	Non-diagnostic but acceptable	Severe but acceptable	Above average
1	Non-diagnostic and unacceptable	Severe and unacceptable	Unacceptable

Statistical analysis

Continuous variables are given as mean ± standard deviation. The Shapiro–Wilk and Kolmogorov–Smirnov tests were used to evaluate normality. The CT attenuation, CNR, SNR, and image noise were compared using one-way analysis of variance (ANOVA) among the three image reconstruction methods (FBP, hybrid IR, and DLR) and three iodine concentration contrast media (Iohexol-370, Iohexol-300, Iohexol-240), followed by a post-hoc test using possible pairwise comparison by a Tukey–Kramer test. Concurrently, a two-sample paired t-test was performed to compare image reconstruction methods and different iodine concentrations of contrast media groups. Subjective image scores were compared using the Wilcoxon signed-rank test. Inter-observer agreement in the qualitative evaluation was assessed with Cohen's kappa coefficient (k), where k < 0.20, 0.31 < k < 0.40, 0.41 < k < 0.60, 0.61 < k < 0.80, and 0.81 < k < 1.00 represent poor, fair, moderate, substantial, and near-perfect agreement, respectively. All statistical analyses were performed using SPSS version 25.0 (IBM Corp., Armonk, NY, USA). Post-hoc power analyses of the results yielded 80.0% power of the test with a sample size of 63, an effect size of 0.4, and an alpha level of 0.05 (G*power version 3.1.9.6, Düsseldorf, Germany). P < 0.05 was considered to represent a statistically significant difference.

Results

Table 2 describes the characteristics of the 63 patients. The image noise was significantly lower in DLR (Iohexol-370: 10.2 ± 2.4; Iohexol-300: 9.1 ± 2.0; Iohexol-240: 9.8 ± 2.2) compared with hybrid IR (Iohexol-370: 23.1 ± 6.0; Iohexol-300: 21.2 ± 4.8; Iohexol-240: 22.6 ± 3.4) and FBP (Iohexol-370: 48.3 ± 6.9; Iohexol-300: 39.2 ± 8.5; Iohexol-240: 43.6 ± 7.9) for all three iodine concentrations (P < 0.001).

Table 2.

Patient characteristics.

Total No. of patients	N = 63
Age (years)	59.32 ± 13.02 (26–86)
Sex (M)	43 (68.25)
BMI (kg/m²)	24.12 ± 3.53
Medical history
Diabetes	13 (20.63)
Hypertension	16 (25.40)
Hyperlipidemia	12 (19.05)
Family history of CVD	4 (6.35)
Smoking
Former smokers	4 (6.35)
Ex-smokers	8 (12.07)

Values are given as n (%) or mean ± SD (range).

BMI, body mass index; CVD, cardiovascular disease.

CT attenuation: Iohexol-370 versus Iohexol-300 versus Iohexol-240

Fig. 1 and Tables 3 and 4 present the scatter plot results of CT attenuation in the left ventricle, coronary arteries, fat, and muscle among the different iodine concentrations (green square = FBP; blue circle = hybrid IR; red diamond = DLR). The mean CT attenuation of the left ventricle was highest among different image reconstructions and iodine concentrations as follows: Iohexol-370 (DLR: 662.9 ± 115.4; hybrid IR: 548.7 ± 83.6; FBP: 550.3 ± 84.7); Iohexol-300 (DLR: 456.1 ± 94.8; hybrid IR: 402.9 ± 66.8; FBP: 404.8 ± 69.2); and Iohexol-240 (DLR: 370.3 ± 69.7; hybrid IR: 336.7 ± 48.8; FBP: 338.9 ± 53.9) (P = 0.136). A post-hoc Tukey's test revealed significant differences (P < 0.001) between each of the different groups (FBP vs. DLR, hybrid IR vs. DLR, Iohexol-240 vs. Iohexol-300, Iohexol-300 vs. Iohexol-370, and Iohexol-240 vs. Iohexol-370). However, no significant difference was found between FBP and hybrid IR (P = 0.558). In addition, the mean CT attenuation value of the coronary artery vessels including LCA, RCA, LAD, and LCX did not show significant differences between Iohexol-370 (DLR: 610.7 ± 98.4; hybrid IR: 505.1 ± 73.9; FBP: 525.0 ± 78.5), Iohexol-300 (DLR: 435.2 ± 84.8; hybrid IR: 384.8 ± 57.5; FBP: 392.2 ± 61.7), and Iohexol-240 (DLR: 352.3 ± 65.1; hybrid IR: 319.5 ± 50.9; FBP: 331.5 ± 55.7) (P = 0.096). Post-hoc Tukey's test resulted in significant differences (P < 0.001) in all possible pairwise comparisons between image reconstructions and iodine concentrations. Therefore, DLR showed significantly higher CT attenuation in the left ventricle and coronary arteries compared with FBP and hybrid IR, regardless of iodine concentration. The CT attenuation of fat (FBP vs. DLR, P = 0.919; hybrid IR vs. DLR, P = 0.995; FBP vs. hybrid IR, P = 0.953) and muscle (FBP vs. DLR, P = 0.974; hybrid IR vs. DLR, P = 0.933; FBP vs. hybrid IR, P = 0.989) was not significantly higher for different iodine concentrations and image reconstructions. Similar to the results of CT attenuation, the histogram plots for the left ventricle show a shift to the right in DLR at different iodine concentrations compared with the images reconstructed with FBP and hybrid IR (Fig. 2). The distribution of attenuation in the histogram plots for fat and muscle was maintained for all image reconstructions at different iodine concentrations. Furthermore, the narrower DLR distribution indicates its lower variability in image noise and CT attenuation compared with FBP and hybrid IR. Notably, DLR boosted CT attenuation in the left ventricle and coronary arteries more than in fat and muscle.

Fig. 1.

CT attenuation results among different iodine concentrations. The mean CT attenuation value of the (a) left ventricle, (b) coronary arteries, (c) fat, and (d) muscle were comparable among different image reconstructions (green square = FBP; blue circle = hybrid IR; red diamond = DLR) and iodine concentrations. Post-hoc Tukey's test resulted in a difference between each group in all three iodine concentrations of contrast media (FBP vs. DLR, P < 0.001; hybrid IR vs. DLR, P < 0.001). For better representation, the comparison of the mean CT attenuation with different image reconstructions is illustrated by the solid line between different iodine concentrations. CT, computed tomography; DLR, deep learning reconstruction; FBP, filtered back projection; IR, iterative reconstruction.

Fig. 2.

Histogram plots with different image reconstructions at various iodine concentrations (iohexol-370, iohexol-300, and iohexol-240). CT attenuation was measured by placing a circular ROI (red square) in the muscle, fat, and left ventricle in axial coronary CTA (representing CTA image, reconstructed with DLR at Iohexol-370, right column). The histogram plots of the left ventricle show a narrow width and shift to the right in DLR (red color) at different iodine concentrations compared with images reconstructed with FBP (green color) and hybrid IR (blue color). The distribution of attenuation in the fat and muscle was maintained for all image reconstructions at different iodine concentrations. Furthermore, a narrower DLR distribution indicates its lower variability in image noise and CT attenuation compared with FBP and hybrid IR. Notably, DLR particularly boosted CT attenuation in the left ventricle and coronary arteries compared to fat and muscle. CTA, computed tomography angiography; DLR, deep learning reconstruction; FBP, filtered back projection; IR, iterative reconstruction; ROI, region of interest.

Table 3.

CT attenuation results at different iodine concentrations.

	FBP (A)	Hybrid IR (B)	DLR (C)	P value
	FBP (A)	Hybrid IR (B)	DLR (C)	A vs. B vs. C	A vs. B	A vs. C	B vs. C
CT attenuation (left ventricle)
Iohexol-240	338.9 ± 53.9	336.7 ± 48.8	370.3 ± 69.7	0.136	0.558	0.001	0.001
Iohexol-300	404.8 ± 69.2	402.9 ± 66.8	456.1 ± 94.8
Iohexol-370	550.3 ± 84.7	548.7 ± 83.6	662.9 ± 115.4
CT attenuation (coronary arteries)
Iohexol-240	331.5 ± 55.7	319.5 ± 50.9	352.3 ± 65.1	0.096	0.001	0.001	0.001
Iohexol-300	392.2 ± 61.7	384.8 ± 57.5	435.2 ± 84.8
Iohexol-370	525.0 ± 78.5	505.1 ± 73.9	610.7 ± 98.4

Values are given as mean ± standard deviation.

CT, computed tomography; DLR, deep learning reconstruction; FBP, filtered back projection; IR, iterative reconstruction.

Table 4.

Comparison of CT attenuation among different iodine concentrations.

P value – CT attenuation of left ventricle
	FBP 240	Hybrid IR 240	DLR 240	FBP 300	Hybrid IR 300	DLR 300	FBP 370	Hybrid IR 370	DLR 370
FBP 240		0.673	*		*	*		*	*
Hybrid IR 240	0.673		*	*		*	*		*
DLR 240	*	*		0.076	0.089		*	*
FBP 300		*	0.076		0.113	*		*	*
Hybrid IR 300	*		0.089	0.113		*	*		*
DLR 300	*	*		*	*		*	*
FBP 370		*	*		*	*		0.160	*
Hybrid IR 370	*		*	*		*	0.160		*
DLR 370	*	*		*	*		*	*
P value – CT attenuation of coronary arteries
	FBP 240	Hybrid IR 240	DLR 240	FBP 300	Hybrid IR 300	DLR 300	FBP 370	Hybrid IR 370	DLR 370
FBP 240		*	*		*	*		*	*
Hybrid IR 240	*		*	*		*	*		*
DLR 240	*	*		*	0.072		*	*
FBP 300		*	*		0.123	*		*	*
Hybrid IR 300	*		0.072	0.123		*	*		*
DLR 300	*	*		*	*		*	*
FBP 370		*	*		*	*		*	*
Hybrid IR 370	*		*	*		*	*		*
DLR 370	*	*		*	*		*	*

*Significant difference (P < 0.05).

CT, computed tomography; DLR, deep learning reconstruction; FBP, filtered back projection; IR, iterative reconstruction.

SNR and CNR: Iohexol-370 versus Iohexol-300 versus Iohexol-240

Fig. 3 presents the bar graph results of the SNR and CNR in the left ventricle and coronary arteries among the different iodine concentrations (green = FBP; blue = hybrid IR; red = DLR). The SNR and CNR in the left ventricle and coronary arteries were significantly higher in DLR compared with FBP and hybrid IR among different iodine concentrations (P < 0.001).

Fig. 3.

The results of CNR and SNR. SNR and CNR in the (a) coronary arteries and (b) left ventricle were significantly higher in DLR (red color) than in FBP (green color) and hybrid IR (blue color) at all concentrations. In particular, DLR with the lowest iodine concentration resulted in significantly higher SNR and CNR compared with FBP and hybrid IR with the highest iodine concentration. The length of the error bars represents the SEM. CNR, contrast-to-noise ratio; DLR, deep learning reconstruction; FBP, filtered back projection; IR, iterative reconstruction; SEM, standard error of the mean; SNR, signal-to-noise ratio.

Subjective analysis: Iohexol-370 versus Iohexol-300 versus Iohexol-240

Table 5 summarizes the subjective image analysis scores (overall image quality, artifacts, and image noise) for image reconstructions with different iodine concentrations. The subjective analysis scores for DLR were significantly higher than those of FBP and hybrid IR with Iohexol-370, Iohexol-300, and Iohexol-240 (P < 0.001). DLR was rated acceptable for image quality with minor artifacts and below-average image noise compared with FBP and hybrid IR for all iodine concentrations. The inter-observer agreement for the image quality between two observers was substantial (k = 0.69).

Table 5.

Results of subjective image analysis.

		FBP	Hybrid IR	DLR	P value
		FBP	Hybrid IR	DLR	FBP vs. hybrid IR	Hybrid IR vs. DLR	FBP vs. DLR
	Iohexol-240
Image quality		3.52 ± 0.60	3.86 ± 0.48	4.28 ± 0.46	0.052	0.001	0.001
Artifact		3.76 ± 0.44	3.86 ± 0.36	4.43 ± 0.60	0.317	0.001	0.001
Image noise		3.24 ± 0.44	3.95 ± 0.38	4.95 ± 0.22	0.001	0.001	0.001
Mean score		3.51 ± 0.26	3.89 ± 0.26	4.55 ± 0.30	0.001	0.001	0.001
	Iohexol-300
Image quality		3.95 ± 0.38	3.76 ± 0.44	4.38 ± 0.67	0.046	0.001	0.001
Artifact		3.33 ± 0.66	3.43 ± 0.59	4.14 ± 0.65	0.317	0.001	0.001
Image noise		3.04 ± 0.67	3.71 ± 0.56	4.52 ± 0.51	0.001	0.001	0.001
Mean score		3.44 ± 0.53	3.63 ± 0.43	4.35 ± 0.45	0.009	0.001	0.001
	Iohexol-370
Image quality		3.95 ± 0.59	4.24 ± 0.43	4.57 ± 0.50	0.014	0.001	0.001
Artifact		3.81 ± 0.40	3.95 ± 0.22	4.47 ± 0.60	0.083	0.001	0.001
Image noise		3.05 ± 0.21	4.00 ± 0.00	4.95 ± 0.21	0.001	0.001	0.001
Mean score		3.60 ± 0.23	4.06 ± 0.30	4.67 ± 0.28	0.001	0.001	0.001

Values are given as mean ± standard deviation.

DLR, deep learning reconstruction; FBP, filtered back projection; IR, iterative reconstruction.

CT attenuation: DLR at Iohexol-300 versus hybrid IR for Iohexol-240

The CT attenuation values in the left ventricle were not significantly different between hybrid IR with Iohexol-300 and DLR with Iohexol-240 (P = 0.089). Moreover, CT attenuation of the coronary arteries was equivalent between hybrid IR with Iohexol-300 and DLR with Iohexol-240 (P = 0.072). Fig. 4 shows an example of coronary CTA image reconstructed with hybrid IR and DLR with the administration of Iohexol-300 and Iohexol-240 contrast media.

Fig. 4.

Visualization of coronary arteries by hybrid IR with Iohexol-300 and DLR with Iohexol-240. (a) For the vessel wall delineation, DLR with Iohexol-240 was scored as very good, while hybrid IR with Iohexol-300 was rated as good (P < 0.001). (b) DLR with lower iodine concentration (Iohexol-240) obtained comparable CT attenuation to hybrid IR with Iohexol-300, whereas higher SNR and CNR were found using DLR. CNR, contrast-to-noise ratio; CT, computed tomography; DLR, deep learning reconstruction; FBP, filtered back projection; IR, iterative reconstruction; SEM, standard error of the mean; SNR, signal-to-noise ratio.

SNR and CNR: DLR at Iohexol-300 versus hybrid IR for Iohexol-240

Although there was no significant difference in the CT attenuation of the left ventricle and coronary arteries between hybrid IR with Iohexol-300 and DLR with Iohexol-240, we discovered that application of DLR with Iohexol-240 significantly improved SNR and CNR in the left ventricle (40.1 ± 13.4 and 33.4 ± 12.9) and coronary arteries (37.9 ± 11.6 and 31.2 ± 11.3) compared with hybrid IR with Iohexol-300 (21.7 ± 4.6; 18.58 ± 4.4 and 20.73 ± 4.08; 17.6 ± 3.8, respectively; P < 0.001) (Fig. 4).

Subjective analysis: DLR at Iohexol-300 versus hybrid IR for Iohexol-240

Besides rating the overall image quality, artifacts, and image noise for image reconstructions at different iodine concentrations, the vessel wall delineation was rated between the results of hybrid IR with Iohexol-300 and DLR with Iohexol-240. The vessel wall delineation is notably higher for DLR with Iohexol-240 groups (score 5 = 8.7%; score 4 = 66.3%; score 3 = 21%; score 2 = 4%) compared with hybrid IR with Iohexol-300 (score 5 = 6.0%; score 4 = 51.3%; score 3 = 38.34%; score 2 = 4.3%), resulting in good vessel wall definition, as shown in Fig. 5. The inter-observer agreement for the visualization of the coronary artery delineation between two observers was substantial (k = 0.76).

Fig. 5.

Subjective image analysis scores between DLR with Iohexol-240 and hybrid IR with Iohexol-300. The coronary arteries were divided into 15 segments, and the vessel wall delineation was rated on a 5-point Likert scale. DLR with Iohexol-240 achieved higher subjective analysis scores for vessel wall delineation compared with hybrid IR with Iohexol-300. DLR, deep learning reconstruction; IR, iterative reconstruction.

Discussion

This study investigated the influence of three different image reconstructions (FBP, hybrid IR, and DLR) on different contrast media iodine concentrations (Iohexol-370, Iohexol-300, and Iohexol-240) in coronary CTA. The mean CT attenuation values of the left ventricle and coronary arteries were significantly higher for DLR compared with FBP and hybrid IR regardless of iodine concentration. In addition, SNR, CNR, and subjective image analysis showed the best results for DLR at all iodine concentrations. In particular, we observed that using DLR with lower iodine concentration (Iohexol-240) yielded comparable CT attenuation to hybrid IR with Iohexol-300, whereas higher SNR, CNR, and subjective analysis scores were found for DLR.

Contrast enhancement in CTA can be effectively achieved with a higher injection rate of contrast media at a lower concentration. Lubbers et al. (26) assessed the intraluminal opacification and image quality between Iodixanol 270 (iso-osmolar contrast media, injection rate 5.6 mL/s) and Iopromide 300 (low-osmolar contrast media, injection rate 5.0 mL/s or 5.8 mL/s) contrast media. They reported that less heat discomfort was associated with the administration of iso-osmolar contrast media at a higher injection rate. Similarly, Mihl et al. (6) concluded that the use of low-iodine-concentration contrast media (contrast media 240 mg/mL at 9 mL/s) with high flow demonstrated no significant differences in the attenuation of coronary arteries compared with contrast media 300 mg/mL at 7.2 mL/s. Consequently, there were no significant differences in the attenuation of the left ventricle and coronary arteries with the administration of Iohexol-240 contrast media for DLR and Iohexol-300 contrast media for hybrid IR at the same flow rate. Additionally, DLR with Iohexol-240 showed better results for SNR, CNR, and subjective image analysis compared with hybrid IR with Iohexol-300. Therefore, DLR is more advantageous to patients with impaired renal function as it requires the administration of lower-iodine-concentration contrast media without increasing the injection rate.

We compared the image quality of coronary CTA among FBP, hybrid IR, and DLR using image noise, SNR, CNR, and subjective image analysis. Previous studies concluded that the image quality of coronary CTA was significantly higher for DLR, with lower image noise, higher SNR, higher CNR, and sharper images compared with FBP and hybrid IR (17). In this study, we investigated the image quality between different image reconstructions besides the administration of various iodine concentrations of contrast media. As shown in Fig. 3, that application of DLR at different iodine concentrations resulted in superior image quality, including higher CNR, higher SNR, and less image noise compared with hybrid IR and FBP. Although DLR at the lowest iodine concentration (Iohexol-240) resulted in significantly lower CT attenuation in the coronary arteries compared with hybrid IR (Iohexol-370), DLR showed better image quality in terms of SNR, CNR, and subjective analysis. This pattern was confirmed by a marked reduction of image noise with DLR at different iodine concentrations compared to FBP and hybrid IR, which led to improved SNR and CNR. Based on the results of this study, the administration of low-iodine-concentration contrast media (Iohexol-240) with an increase in flow rate of contrast media in DLR could achieve comparable CT attenuation to higher-iodine-concentration (Iohexol-370) hybrid IR. We believe this clear contrast enhancement in the left ventricle and coronary arteries with DLR is due to the advanced MBIR training images. Several studies have shown a marked improvement of vascular contrast enhancement with MBIR compared with FBP and IRs (22,27).

Previous studies investigated low tube voltage protocols to achieve better CT attenuation with lower radiation dose. However, this method is not appropriate for larger patient datasets (28). Moreover, the presence of calcification in the vessels and higher blooming artifacts caused blurring of the coronary vessels, which led to misdiagnosis and streak artifacts in the low tube voltage (28,29). This study demonstrated that DLR with the lowest iodine concentration (Iohexol-240) resulted in significantly higher SNR and CNR compared with hybrid IR with the highest iodine concentration (Iohexol-370) at the same flow rate and tube voltage. Furthermore, it has been reported that DLR provides lower image noise at lower tube voltage (30). In addition, the blooming artifact reduction is significantly improved with DLR compared with FBP and hybrid IR (21). These advantages could be applied to DLR with a lower radiation dose scanning protocol. In addition, the contrast enhancement boost is a newly introduced technique designed to improve the strength of contrast enhancement. This technique has been applied to the visualization of Type-II endoleaks in abdominal CTA (31). The diagnostic value of CTA directly depends on the degree of vascular enhancement (32). Thus, further work is required to investigate the effect of the contrast enhancement boost technique on various iodine concentrations of contrast media with DLR in coronary CTA.

The present study has some limitations. First, we did not assess hemodynamic changes (heart rate, blood pressure, and patient discomfort) during the administration of contrast media. The injection rate of contrast media was the same for all iodine concentrations. Previous studies examined pain, heat, and discomfort with higher flow rate and iodine concentrations of contrast media (26). Nevertheless, patients with renal impairment who receive lower iodine concentrations would experience less discomfort and toxicity. Second, this is a single-center study with a small sample size. Further studies are required to evaluate the diagnostic performance among various iodine concentrations with DLR. Third, all investigations were only performed on Aquilion ONE Prism CT vendor. The effect of various image reconstructions with different kernels and options for image quality could be different depending on the CT vendor, and this may have affected the results of this study.

In conclusion, we suggest that using DLR with Iohexol-240 contrast media is preferred over hybrid IR with Iohexol-300 contrast media in coronary CTA.

Footnotes

Declaration of conflicting interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: JKR and HJS are employees of Canon Medical Systems Korea, Seoul, Republic of Korea (the subsidiary in Korea of Canon Medical Systems Corporation, Otawara-si, Japan).

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Korea Medical Device Development Fund grant funded by the Korea government (the Ministry of Science and ICT, the Ministry of Trade, Industry and Energy, the Ministry of Health & Welfare, the Ministry of Food and Drug Safety) (project no. 1711139017).

ORCID iD

Chuluunbaatar Otgonbaatar

References

Weininger

Barraza

Kemper

, et al. Cardiothoracic CT angiography: current contrast medium delivery strategies. AJR Am J Roentgenol 2011;196:W260–W272.

Flohr

Schoepf

Ohnesorge

. Chasing the heart: new developments for cardiac CT. J Thorac Imaging 2007;22:4–16.

Schoepf

. Contrast media for coronary CT angiography: should an iso-osmolar agent be used? Radiology 2018;286:81–82.

Awai

Hiraishi

Hori

. Effect of contrast material injection duration and rate on aortic peak time and peak enhancement at dynamic CT involving injection protocol with dose tailored to patient weight. Radiology 2004;230:142–150.

Kim

Cho

Lee

, et al. Low-tube-voltage CT urography using low-concentration-iodine contrast Media and iterative reconstruction: a multi-institutional randomized controlled trial for comparison with conventional CT urography. Korean J Radiol 2018;19:1119–1129.

Mihl

Kok

Wildberger

, et al. Coronary CT angiography using low concentrated contrast media injected with high flow rates: feasible in clinical practice. Eur J Radiol 2015;84:2155–2160.

Lin

Zhou

Schoepf

, et al. Tube voltage, DNA double-strand breaks, and image quality in coronary CT angiography. Korean J Radiol 2020;21:967–977.

Nakaura

Awai

Maruyama

, et al. Abdominal dynamic CT in patients with renal dysfunction: contrast agent dose reduction with low tube voltage and high tube current-time product settings at 256-detector row CT. Radiology 2011;261:467–476.

Van Cauteren

Van Gompel

Tanaka

, et al. The impact of combining a low-tube voltage acquisition with iterative reconstruction on total iodine dose in coronary CT angiography. Biomed Res Int 2017;2017:2476171.

10.

Sultana

Morishita

Awai

, et al. Evaluation of hypervascular hepatocellular carcinoma in cirrhotic liver by means of helical CT: comparison of different contrast medium concentrations within the same patient. Radiat Med 2003;21:239–245.

11.

Tsurusaki

Sugimoto

Fujii

, et al. Multi-detector row helical CT of the liver: quantitative assessment of iodine concentration of intravenous contrast material on multiphasic CT–a prospective randomized study. Radiat Med 2004;22:239–245.

12.

Ding

Meystre

Campeanu

, et al. Contrast media extravasations in patients undergoing computerized tomography scanning: a systematic review and meta-analysis of risk factors and interventions. JBI Database System Rev Implement Rep 2018;16:87–116.

13.

Gruberg

Mintz

Mehran

, et al. The prognostic implications of further renal function deterioration within 48 h of interventional coronary procedures in patients with pre-existent chronic renal insufficiency. J Am Coll Cardiol 2000;36:1542–1548.

14.

Boedeker

. AiCE Deep Learning Reconstruction: Bringing the Power of Ultra-High Resolution CT to Routine Imaging. 2017. Available at: https://www.global.medical.canon/publication/ct/2019WP_AiCE_Deep_Learning (accessed 20 January 2022).

15.

Hsieh

Liu

Nett

, et al. A new era of image reconstruction: TrueFidelity – technical white paper on deep learning image reconstruction. 2019. Available at: https://www.gehealthcare.com/-/jssmedia/040dd213fa89463287155151fdb01922.pdf (accessed 12 January 2022).

16.

Akagi

Nakamura

Higaki

, et al. Deep learning reconstruction improves image quality of abdominal ultra-high-resolution CT. Eur Radiol 2019;29:6163–6171.

17.

Tatsugami

Higaki

Nakamura

, et al. Deep learning-based image restoration algorithm for coronary CT angiography. Eur Radiol 2019;29:5322–5329.

18.

Lenfant

Chevallier

Comby

, et al. Deep learning versus iterative reconstruction for CT pulmonary angiography in the emergency setting: improved image quality and reduced radiation dose. Diagnostics (Basel 2020; 10: 558.

19.

Kim

Kang

Yoon

, et al. Deep learning-based image reconstruction for brain CT: improved image quality compared with adaptive statistical iterative reconstruction-Veo (ASIR-V). Neuroradiology 2021;63:905–912.

20.

Bernard

Comby

Lemogne

, et al. Deep learning reconstruction versus iterative reconstruction for cardiac CT angiography in a stroke imaging protocol: reduced radiation dose and improved image quality. Quant Imaging Med Surg 2021;11:392–401.

21.

Otgonbaatar

Ryu

J-K

Kim

, et al. Improvement of depiction of the intracranial arteries on brain CT angiography using deep learning reconstruction. J Integr Neurosci 2021;20:967–976.

22.

Hamamura

Hayashida

Takeshita

, et al. The usefulness of full-iterative reconstruction algorithm for the visualization of cystic artery on CT angiography. Jpn J Radiol 2019;37:526–533.

23.

Nebelung

Brauer

Seppelt

, et al. Coronary computed tomography angiography (CCTA): effect of bolus-tracking ROI positioning on image quality. Eur Radiol 2021;31:1110–1118.

24.

Shi

Han

, et al. Image quality and radiation dose for prospectively triggered coronary CT angiography: 128-slice single-source CT versus first-generation 64-slice dual-source CT. Sci Rep 2016;6:34795.

25.

Austen

Edwards

Frye

, et al. A reporting system on patients evaluated for coronary artery disease. Report of the Ad Hoc committee for grading of coronary artery disease, council on cardiovascular surgery, American Heart Association. Circulation 1975;51:5–40.

26.

Lubbers

Kock

Niezen

, et al. Iodixanol versus iopromide at coronary CT angiography: lumen opacification and effect on heart rhythm-the randomized IsoCOR trial. Radiology 2018;286:71–80.

27.

Katsura

Sato

Akahane

, et al. Comparison of pure and hybrid iterative reconstruction techniques with conventional filtered back projection: image quality assessment in the cervicothoracic region. Eur J Radiol 2013;82:356–360.

28.

Seyal

Arslanoglu

Abboud

, et al. CT of the abdomen with reduced tube voltage in adults: a practical approach. Radiographics 2015;35:1922–1939.

29.

Lee

Kim

Lee

, et al. Diagnostic accuracy of low-radiation coronary computed tomography angiography with low tube voltage and knowledge-based model reconstruction. Sci Rep 2019;9:1308.

30.

Wang

Fan

Shi

, et al. A deep-leaning reconstruction algorithm that improves the image quality of low-tube-voltage coronary CT angiography. Eur J Radiol 2021;146:110070.

31.

Iizuka

Yokota

Kidoh

, et al. Contrast enhancement boost technique at aortic computed tomography angiography: added value for the evaluation of type II endoleaks after endovascular aortic aneurysm repair. Acad Radiol 2019;26:1435–1440.

32.

Swanberg

Nyman

Magnusson

, et al. Selective intra-arterial dual-energy CT angiography (s-CTA) in lower extremity arterial occlusive disease. Eur J Vasc Endovasc Surg 2014;48:325–329.