Translate this page into:
Comparison of image quality in carotid dual-energy computed tomography angiography at 55 keV virtual monoenergetic imaging using deep learning and adaptive iterative reconstruction algorithm
*Corresponding author: Yankai Meng, Department of Radiology, the Affiliated Hospital of Xuzhou Medical University Xuzhou,China. mengyankai@126.com
-
Received: ,
Accepted: ,
How to cite this article: Liu X, Wang C, Long J, Wu Y, Liu Z, Yu M, et al. Comparison of image quality in carotid dual-energy computed tomography angiography at 55 keV virtual monoenergetic imaging using deep learning and adaptive iterative reconstruction algorithm. J Clin Imaging Sci. 2025;15:42. doi: 10.25259/JCIS_109_2025
Abstract
Objectives:
This study aims to evaluate the image quality of 55 keV virtual monoenergetic imaging (VMI) in carotid dual-energy computed tomography (CT) angiography (DE-CTA) reconstructed using deep learning image reconstruction (DLIR) algorithms and traditional iterative reconstruction algorithms.
Material and Methods:
This prospective study included 48 patients who underwent DE-CTA examinations at our institution between December 2024 and January 2025. Image reconstructions were performed using 50% strength adaptive statistical iterative reconstruction-Veo (ASIR-V 50%), low and high strengths DLIR (DLIR-L and DLIR-H) algorithms. Objective image quality was evaluated by measuring background noise (standard deviation), signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) at key anatomical locations, including the aortic arch, common carotid artery, carotid bifurcation, and internal carotid artery. Two senior radiologists conducted subjective assessments of image quality, focusing on image noise, artifacts, and vessel continuity, and the clarity of vascular wall margin.
Results:
Compared with ASIR-V 50% and DLIR-L, DLIR-H significantly improved image quality by reducing background noise and increasing SNR and CNR (P < 0.05). Subjectively, DLIR-H images demonstrated better suppression of noise and clearer vascular wall margin (P < 0.05). Subgroup analysis revealed that these improvements were more pronounced in patients with a body mass index (BMI) ≥24 kg/m2. No significant differences were observed in CT attenuations among the three reconstruction methods (P > 0.05).
Conclusion:
At 55 keV VMI in carotid DE-CTA, DLIR-H significantly enhanced image quality, particularly by reducing noise and preserving fine anatomical structures. Its efficacy was especially notable in patients with BMI ≥24 kg/m2.
Keywords
Carotid computed tomography angiography
Deep learning
Dual-energy computed tomography
Image quality
Iterative reconstruction
INTRODUCTION
Acute cerebrovascular events were responsible for approximately 7.3 million deaths in 2021, ranking as the third leading cause of mortality worldwide.[1] One of the primary triggers of these events is the rupture of carotid artery plaques. The morphological and functional characteristics of these plaques are essential for accurate diagnosis, risk prognosis, and therapeutic planning.[2]
Head-and-neck computed tomography angiography (CTA) is a widely used imaging technique and the current standard for evaluating carotid plaques.[3] Compared with conventional single-energy computed tomography (CT), dual-energy CTA (DE-CTA) offers enhanced characterization of plaque components, such as lipid cores and markers of plaque vulnerability.[4,5] Low-energy virtual monoenergetic imaging (VMI), derived from dual-energy data, has been shown to enhance vessel contrast, reduces the required volume of contrast agents, and improves diagnostic performance.[6,7] Studies, including those by Zhou et al., had identified 55 keV VMI as the optimal setting for CTA, achieving comparable image quality while minimizing both contrast agent dosage and radiation doses.[8]
Although adaptive statistical iterative reconstruction-Veo (ASIR-V) has been shown to improve image quality under low-contrast and low-radiation conditions,[9,10] it can introduce excessive smoothing, potentially masking fine anatomical structures. In contrast, deep learning-based image reconstruction (DLIR) algorithm has emerged as a powerful alternative, capable of significantly reducing image noise while preserving anatomical details.[11-13] Li et al. reported that high-strength DLIR (DLIR-H) outperformed ASIR-V in terms of image quality in low-radiation coronary CTA.[14] Similarly, Jiang et al. observed enhanced image quality in low-keV VMI reconstructions using DLIR.[15] However, the performance of DLIR in enhancing low-keV VMI images under conditions of both reduced contrast agent and radiation doses (dual-low) – especially in obese patients (body mass index [BMI] >24 kg/m2) – has yet to be fully elucidated.
Therefore, this study aims to compare the image quality of 55 keV VMI DE-CTA reconstructions generated using DLIR and ASIR-V under such “dual-low” imaging conditions.
MATERIAL AND METHODS
Study population
This prospective study included 48 patients who underwent DE-CTA examinations at our institution between December 2024 and January 2025. Written informed consent was obtained from all patients, and the study protocol was approved by the Institution Ethics Committee (Ethical approval number: XYFY2024–KL456). Inclusion criteria: Allergy to iodine contrast agents; severe cardiac, hepatic, or renal dysfunction, including patients with estimated glomerular filtration rate <30 mL/min/1.73 m2, or other conditions that may affect the contrast agent metabolism or imaging accuracy; hyperthyroidism; pregnancy or breastfeeding status; inability to cooperate due to involuntary movement or cognitive impairment that would hinder adequate image acquisition or interpretation; and severe claustrophobia or other contraindications to CT imaging [Figure 1].
- Flowchart of this study. (DE-CTA: Dual-Energy Carotid Artery CT Angiography, ASIR-V: Adaptive Statistical Iterative Reconstruction-V, DLIR-L: Deep Learning Image Reconstruction with Low Setting; DLIR-H: Deep Learning Image Reconstruction with High Setting, GSI: Gemstone Spectral Imaging, VMI: Virtual Monoenergetic Imaging, CT: computed tomography, SD: standard deviation, SNR: Signal-to-noise ratio, CNR: contrast-to-noise ratio)
CT image acquisition and reconstruction
All patients were performed on a 256-row CT (Revolution APEX Expert, GE Healthcare, USA). Scanning was performed from the level of the tracheal bifurcation to the cranial vertex. Smart prep tracking was used with the region of interest (ROI) monitored at the level of the ascending aorta at the aortic arch (AOA) level. The trigger threshold was set at 50 HU, with a delay time of 2 s. Iodixanol (350 mgI/mL, Hengrui Medicine, China) injected through a high-pressure injector (CT Motion, Olyric Medical Co., Germany) through the median cubital vein. The injection dose was 0.5 mL/kg body weight, completed within 10 s, followed by an injection of 40 mL saline.
The DE-CTA scanning parameters were as follows. The gemstone spectral imaging (GSI) mode was employed, with the tube voltage alternating between 80 kV and 140 kV. The tube current was controlled by GSI Assist, with a noise index of 4.0. The pitch was 0.992:1, the scanning speed was 0.5 s/r, and the detector width was 128 × 0.625 mm. ASIR-V 50%, DLIR-L, and DLIR-H were used to reconstruct the raw data into a spectral image dataset (GSI Data File), with a reconstruction slice thickness of 1.25 mm, an slice spacing of 0.625 mm, and an image matrix of 512 × 512. The spectral image dataset was then transferred to a GE Advantage Workstation 4.7 (GE, Milwaukee, USA) for post-processing to generate a 55 keV VMI.
Image quality objective evaluation
A radiologist (Y.K.M.) with 15 years of experience in cardiovascular imaging conducted a blinded, randomized analysis of image quality using the picture archiving and communication system (PACS). Measurements were performed at four key anatomical locations: The AOA, the origin of the common carotid artery (CCA), the carotid bifurcation (CCB), and the origin of the internal carotid artery (ICA). All measurements were taken from axial CTA images with a slice thickness of 1.25 mm. The radiologist adjusted the window width and window level as needed. The ROI was positioned at the center of the measured vessel, covering half of its diameter while avoiding calcifications, significant stenosis, or vascular occlusions. The following parameters were recorded: The CT attenuation (CTvessel) and standard deviation (SDvessel) of each vessel, as well as the CT attenuation (CTmuscle) and standard deviation (SDmuscle) of the adjacent sternocleidomastoid muscle. SDMs were considered to be the background noise of the image. Signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were calculated using predefined formulas.
SNR = CTvessel/SDvessel
CNR = (CTvessel-CTmuscle)/SDvessel
Image quality subjective evaluation
Two senior radiologists independently assessed image quality (Y.K.M. with 15 years of experience in diagnostic radiology and K.X. with 20 years of experience). Their average scores were used for statistical analysis. Images from eligible cases were randomly selected from the PACS system by a 2nd-year radiology resident, who was also responsible for data collection. The senior radiologists were blinded to scanning parameters, reconstruction methods, and demographic characteristics of the included cases. Image quality assessment was conducted in the same room under consistent lighting conditions, using identical dedicated monitors. The senior radiologists reviewed the original axial images and multiplanar reconstruction images sequentially, with the option to adjust window width and window level as needed. The observation time for each case was at least 5 min.
Subjective scoring was based on five criteria: Image noise, artifacts, vascular continuity, vascular wall delineation, and diagnostic confidence. Each criterion was rated on a 5-point scale, with detailed definitions provided in the Supplementary Material Table 1. A 2nd-year radiology resident calculated the total score based on the senior radiologists’ ratings. The overall image quality was classified as follows : 0–5 points: Poor; 6–10 points: Fair; 11–15 points: Moderate; 16–20 points: Good; 21–25 points: Excellent.
Radiation dose
The CT dose index (CTDIvol) and dose-length product of all enrolled patients were recorded for the purpose of calculating the radiation dose. The effective radiation dose (ED) for DE-CTA will be calculated using the formula, where k = 0.0031 mSv/(mGy•cm):
ED = DLP × k
Statistical analysis
Statistical analysis was conducted using the Statistical Package for the Social Sciences, version 26.0. Normally distributed continuous variables were expressed as means ± standard deviation, while non-normally distributed data were reported as the median (interquartile range). Categorical data were expressed as proportions or percentages. The Shapiro-Wilk test was used to assess normality. For normally distributed data with homogeneity of variance, one-way analysis of variance was performed, followed by the least significant difference test for post hoc pairwise comparisons. For data violating normality or homogeneity assumptions, the Kruskal-Wallis test was applied, followed by Dunn’s test for pairwise comparisons. P < 0.05 was considered statistically significant.
RESULTS
Patient demographics
A total of 48 patients were included in this study, comprising 26 males (54.17%) and 22 females (45.83%). The mean age of the patients was 59.92 ± 10.58 years. The average BMI was 25.37 ± 3.45 kg/m2, with 31 patients (64.59%) having a BMI ≥24 kg/m2. The mean contrast agent volume used was 34.03 ± 5.85 mL, and the mean injection rate was 3.40 ± 0.58 mL/s. The average ED value was 1.54 ± 0.09 mSv [Table 1].
| Variable | Value |
|---|---|
| Gender (Male (%)) | 26 (54.2%) |
| Age (years) | 59.92±10.58 |
| Height (m) | 1.64±0.07 |
| Weight (kg) | 68.06±11.70 |
| BMI (kg/m2) | 25.37±3.45 |
| BMI<18.5 | 1 (2.08%) |
| BMI: 18.5-23.9 | 16 (33.33%) |
| BMI≥24 | 31 (64.58%) |
| Contrast Agent Volume (mL) | 34.03±5.85 |
| Contrast Injection Rate (mL/s) | 3.40±0.58 |
| CTDIvol(mGy) | 12.18±0.46 |
| DLP (mGy•cm) | 494.67±29.13 |
| ED (mSv) | 1.54±0.09 |
Note: Data are presented as mean±standard deviation or number (%). BMI: Body mass index, CTDIvol: Volume computed tomography dose index, dLP: Dose length product, ED: Effective dose
Objective image quality evaluation
Across all four evaluated vessels, background noise progressively decreased from the ASIR-V 50% group to the DLIR-L group and was lowest in the DLIR-H group, while both SNR and CNR increased in the same order. No statistically significant differences were found between ASIR-V 50% and DLIR-L (all P > 0.05); however, both differed significantly from DLIR-H (all P < 0.05). Comparisons of CT attenuations among the three reconstruction algorithms revealed no significant differences (all P > 0.05) [Table 2, Figures 2 and 3a-l].
- Comparison of objective image quality assessment metrics, including standard deviation, signal-to-noise ratio and contrast-to-noise ratio across three groups at varying anatomical levels. (ASIR-V: Adaptive statistical iterative reconstruction-veo, DLIR: Deep learning image reconstruction, CT: Computed tomography, SD: Standard deviation, SNR: Signal-to-noise ratio, CNR: Contrast-to-noise ratio, GSI: Gemstone spectral imaging, VMI: Virtual monoenergetic imaging, AOA: Aortic arch, CCA: Common carotid artery, CCB: Carotid bifurcation, ICA: Internal carotid artery).
- A 54-year-old female (BMI 23.88 kg/m2) underwent dual-energy CT angiography (DECTA), reconstructed using (a, d, g, j) 50% ASIR-V, (b, e, h, k) DLIR-L, and (c, f, i, l) DLIR-H. The white arrow indicates the target vessel. (CCA: Common carotid artery, CCB: Carotid bifurcation, ICA: Internal carotid artery. BMI: Body mass index).
| Variable | ASIR-V 50% | DLIR-L | DLIR-H | P-value | P1-value | P2-value | P3-value |
|---|---|---|---|---|---|---|---|
| CT | |||||||
| AOA | 501.863±73.341 | 501.973±73.358 | 502.008±73.241 | 0.999 | 0.931 | 0.928 | 0.853 |
| CCA | 473.633±82.674 | 474.973±82.425 | 473.312±82.727 | 0.985 | 0.916 | 0.809 | 0.922 |
| CCB | 503.104±84.218 | 503.346±84.000 | 503.844±83.797 | 0.983 | 0.907 | 0.856 | 1.000 |
| ICA | 491.775±82.926 | 492.719±82.692 | 493.617±82.676 | 0.974 | 0.898 | 0.844 | 0.988 |
| Background Noise (SD values of the muscle in the same level) | |||||||
| AOA | 25.271±7.076 | 23.979±6.927 | 16.625±6.083 | 0.000 | 0.336 | 0.000 | 0.000 |
| CCA | 25.250±5.401 | 23.271±5.584 | 15.979±4.393 | 0.000 | 0.062 | 0.000 | 0.000 |
| CCB | 11.375±3.456 | 10.042±2.843 | 6.458±2.269 | 0.000 | 0.060 | 0.000 | 0.000 |
| ICA | 10.562±2.211 | 9.438±1.712 | 5.958±1.254 | 0.000 | 0.069 | 0.000 | 0.000 |
| SNR | |||||||
| AOA | 15.280±3.106 | 14.962±3.283 | 19.644±5.441 | 0.000 | 0.580 | 0.000 | 0.000 |
| CCA | 18.221±5.934 | 17.377±5.614 | 21.783±8.330 | 0.020 | 0.551 | 0.039 | 0.008 |
| CCB | 37.399±12.317 | 37.906±10.677 | 50.928±17.955 | 0.000 | 0.984 | 0.000 | 0.000 |
| ICA | 35.529±9.555 | 35.307±9.267 | 45.948±14.821 | 0.000 | 0.876 | 0.000 | 0.000 |
| CNR | |||||||
| AOA | 18.832±5.940 | 19.778±5.471 | 29.764±11.100 | 0.000 | 0.353 | 0.000 | 0.000 |
| CCA | 17.444±5.777 | 19.097±6.181 | 28.014±9.251 | 0.000 | 0.150 | 0.000 | 0.000 |
| CCB | 40.984±13.393 | 46.215±14.397 | 73.555±23.378 | 0.000 | 0.111 | 0.000 | 0.000 |
| ICA | 41.849±11.307 | 46.435±11.459 | 74.539±18.906 | 0.000 | 0.149 | 0.000 | 0.000 |
ASIR-V 50%, 50% strength adaptive statistical iterative reconstruction-Veo algorithm, DLIR-L: Low strength deep learning image reconstruction, DLIR-H: High strength deep learning image reconstruction, AOA: Aortic arch, CCA: Common carotid artery, CCB: Carotid bifurcation, ICA: Internal carotid artery, CT: Computed tomography, SD: Standard deviation, SNR: Signal-to-noise ratio, CNR: Contrast-to-noise ratio, P: intra-group statistics, P1: compares the statistical results between ASIR-V 50% and DLIR-L groups, P2: compares the statistical results between ASIR-V 50% and DLIR-H groups, P3: compares the statistical results between DLIR-L and DLIR-H groups.
Objective image quality evaluation in the BMI ≥24 kg/m2 subgroup
In the subgroup with BMI ≥24 kg/m2, the image quality exhibited a consistent pattern across the four evaluated vessels: Background noise progressively decreased from ASIR-V 50% to DLIR-L and then to DLIR-H, while both the SNR and CNR correspondingly increased. There was no statistically significant difference between ASIR-V 50% and DLIR-L (all P > 0.05); however, both differed significantly from DLIR-H (all P < 0.05). No statistically significant differences in CT values were observed among the three reconstruction algorithms in any pairwise comparisons (all P > 0.05).
Subjective image quality evaluation
In terms of subjective evaluation, DLIR-H demonstrated significantly lower image noise compared to both ASIR-V 50% and DLIR-L (all P < 0.05). Moreover, the DLIR-H group exhibited markedly better delineation of the vascular wall (all P < 0.05). The overall subjective image quality score for DLIR-H was also significantly higher than those of ASIR-V 50% and DLIR-L (all P < 0.05). No statistically significant differences were observed among the three reconstruction algorithms in terms of image artifacts, vascular continuity, or diagnostic confidence (all P > 0.05) [Table 3 and Figure 4ab].
| Variables | ASIR-V 50% | DLIR-L | DLIR-H | P | P1 | P2 | P3 |
|---|---|---|---|---|---|---|---|
| Image noise | 3.319±0.471 | 3.383±0.491 | 3.936±0.323 | 0.000 | 0.478 | 0.000 | 0.000 |
| Image artifacts | 3.787±0.414 | 3.787±0.414 | 3.894±0.312 | 0.301 | 1.000 | 0.180 | 0.180 |
| Vascular continuity | 4.085±0.282 | 4.085±0.282 | 4.043±0.204 | 0.655 | 1.000 | 0.427 | 0.427 |
| Vessel wall margin | 3.617±0.491 | 3.787±0.414 | 4.043±0.292 | 0.000 | 0.045 | 0.000 | 0.003 |
| Diagnostic confidence | 3.958±0.204 | 3.958±0.204 | 3.958±0.204 | 1.000 | 1.000 | 1.000 | 1.000 |
| Total points | 18.766±1.146 | 19.000±1.123 | 19.872±0.741 | 0.000 | 0.268 | 0.000 | 0.000 |
ASIR-V 50%, 50% strength adaptive statistical iterative reconstruction-Veo algorithm, DLIR-L: Low strength deep learning image reconstruction, DLIR-H: High strength deep learning image reconstruction, P: intra-group statistics, P1: compares the statistical results between ASIR-V 50% and DLIR-L groups, P2: compares the statistical results between ASIR-V 50% and DLIR-H groups, P3: compares the statistical results between DLIR-L and DLIR-H groups.
- Comparison of subjective image quality evaluation scoring among three groups. P-values were calculated using one-way ANOVA with LSD post hoc tests for normally distributed data, or the Kruskal-Wallis test with Dunn’s post hoc tests for non-normal data. A significance level of P < 0.05 was used. ASIR-V: Adaptive Statistical Iterative Reconstruction-V, DLIR-L: Deep Learning Image Reconstruction with Low Setting; DLIR-H: Deep Learning Image Reconstruction with High Setting.
DISCUSSION
This study shows that under dual-low scanning conditions in CTA, 55keV VMI processed using the DLIR-H algorithm provide significantly better image quality than both the ASIR-V 50% and DLIR-L algorithms. The DLIR-H algorithm achieved the best SNR and CNR for carotid artery imaging, which is consistent with previous studies.
The traditional ASIR-V algorithm reduces image noise through multiple iterative reconstructions, which can lead to over-smoothing of the images and obscure subtle lesions, making it challenging for evaluating plaque morphology in the vessel wall. In contrast, the DLIR-H algorithm, based on deep convolutional neural networks, reduces noise by training the model on large low-dose datasets, while effectively preserving critical anatomical details and image textures.[16] In this study, subjective evaluations showed that DLIR-H achieved the highest scores for noise reduction and vessel wall clarity, further confirming the advantages of the DLIR algorithm. This is beneficial for better visualization of the carotid artery wall structure and provides high-quality image foundation for precise carotid artery measurements and image segmentation.[17]
A study by Jiang et al. compared the image quality of DLIR-H and ASIR-V in low-keV VMI images, demonstrating that DLIR-H had a greater advantage at the AOA and CCA levels.[15] Building on Jiang et al.’s[15] research, we further reduced the contrast agent concentration (350 mgI/mL) and contrast volume (average reduction of 15%), while the DLIR-H group still achieved the highest SNR and CNR. This allows us to benefit from the contrast advantages of low-keV VMI while mitigating the inherent high noise levels typically associated with low-energy imaging.
In patients with BMI ≥24 kg/m2, DLIR-H also showed superior image quality. Obese or overweight patients face higher risks of atherosclerosis due to mechanisms such as lipid metabolism disorders, hypertension, inflammation, and insulin resistance.[1,18] In addition, these patients tend to have greater tissue attenuation and higher image noise, resulting in poorer image quality. In clinical practice, DLIR-H may provide an optimized solution for these patients, improving diagnostic accuracy.
In this study, we selected four different anatomical levels, the AOA, CCA origin, CCB, and ICA, to assess the overall image quality of DE-CTA. Due to hemodynamic factors, atherosclerotic plaques in the carotid artery are most likely to occur at the vessel origin and bifurcation levels. These four selected levels represent the regions with the highest risk for plaque formation in clinical practice. Under the condition of low contrast agent usage, the image quality of DLIR-H was significantly superior to that of ASIR-V 50% and DLIR-L. We believe that this result can be applied to the clinical evaluation of carotid artery plaques, enhancing the accuracy of plaque analysis. However, the generalizability of this result requires validation through multi-center studies with larger sample sizes.
While this study focused on evaluating DE-CTA with DLIR-H, comparing it with established imaging techniques such as magnetic resonance angiography (MRA) and digital subtraction angiography (DSA) would provide further insights into its clinical applicability. DSA remains the gold standard due to its high resolution, while MRA offers a non-invasive alternative without radiation. Future studies comparing DE-CTA’s diagnostic accuracy, sensitivity, and specificity to DSA and MRA are crucial for assessing its potential as a viable alternative, especially for patients with contraindications to these modalities. Although such comparisons were not included in the present study, they represent an important direction for future research to further establish DE-CTA’s role in carotid artery disease diagnosis.
This study demonstrates that DLIR-H significantly improves image quality in 55 keV VMI carotid DE-CTA, particularly by reducing noise and preserving fine anatomical details. This is especially beneficial in patients with a BMI ≥24 kg/m2, who often experience higher image noise due to increased tissue attenuation. DLIR-H provides clearer images, aiding in more accurate plaque characterization and enhancing diagnostic confidence. In addition, DLIR-H allows for high-quality imaging with reduced contrast agent and radiation doses, improving patient safety, particularly in those with compromised renal function. By integrating DLIR-H into routine clinical practice, clinicians can achieve enhanced diagnostic accuracy while minimizing patient exposure to contrast and radiation.
This study has some limitations. First, the average effective dose in this study was 1.53 ± 0.09 mSv, which still offers room for further reduction compared to previous studies.[19] Second, although we analyzed the subgroup of patients with BMI ≥24 kg/m2, the potential for DLIR to optimize VMI images in patients with BMI ≥28 kg/m2 requires further investigation. Finally, as a single-center clinical study, further multi-center external validation is required.
CONCLUSION
DLIR-H significantly improves image quality in DE-CTA at 55keV VMI, particularly in noise reduction and anatomical detail preservation, when compared to both ASIR-V and DLIR-L algorithms. In addition, DLIR-H was particularly effective in improving image quality in obese patients (BMI ≥24 kg/m2).
Ethical approval:
The research/study was approved by the Institutional Review Board at Xuzhou Medical University Affiliated Hospital Ethics Committee, number XYFY2024-KL456, dated 2024-01.
Declaration of patient consent:
The authors certify that they have obtained all appropriate patient consent.
Conflicts of interest:
There are no conflicts of interest.
Use of artificial intelligence (AI)-assisted technology for manuscript preparation:
The authors confirm that there was no use of artificial intelligence (AI)-assisted technology for assisting in the writing or editing of the manuscript and no images were manipulated using AI.
Financial support and sponsorship: Jiangsu provincial health commission elderly health research project (LKM2022018); Xuzhou health commission technology projects (XWKYHT20220108, XWKYHT20230086, and XWKYHT20240086); affiliated hospital of Xuzhou Medical University Institutional Research Project (2024ZY08)
References
- Global, regional, and national burden of stroke and its risk factors 1990-2021: A systematic analysis for the global burden of disease study 2021. Lancet Neurol. 2024;23:973-1003.
- [Google Scholar]
- Symptomatic nonstenotic carotids: A topical review. Stroke. 2024;55:2921-31.
- [CrossRef] [PubMed] [Google Scholar]
- Correction to: State-of-the-art CT and MR imaging and assessment of atherosclerotic carotid artery disease: The reporting-a consensus document by the European Society of Cardiovascular Radiology (ESCR) Eur Radiol. 2023;33:1499-500.
- [CrossRef] [PubMed] [Google Scholar]
- Utility of dual-layer spectral detector CTA to characterize carotid atherosclerotic plaque components: An imaging-histopathology comparison in patients undergoing endarterectomy. Am J Roentgenol. 2022;218:517-25.
- [CrossRef] [PubMed] [Google Scholar]
- Application of dual-layer spectral-detector computed tomography angiography in identifying symptomatic carotid atherosclerosis: A prospective observational study. J Am Heart Assoc. 2024;13:e032665.
- [CrossRef] [PubMed] [Google Scholar]
- Dual energy CT in clinical routine: How it works and how it adds value. Emerg Radiol. 2021;28:103-17.
- [CrossRef] [PubMed] [Google Scholar]
- Dual-energy computed tomography: Technological considerations. Radiol Clin Nam. 2023;61:945-61.
- [CrossRef] [PubMed] [Google Scholar]
- Low dose iodinated contrast material and radiation for virtual monochromatic imaging in craniocervical dual-layer spectral detector computed tomography angiography: A prospective and randomized study. Acad Radiol. 2024;31:2501-10.
- [CrossRef] [PubMed] [Google Scholar]
- Combination of adaptive statistical iterative reconstruction-V and lower tube voltage during craniocervical computed tomographic angiography yields better image quality with a reduced radiation dose. Acad Radiol. 2019;26:e233-40.
- [CrossRef] [PubMed] [Google Scholar]
- Impact of 80 kVp with iterative reconstruction algorithm and low-dose contrast medium on the image quality of craniocervical CT angiography. Clin Imaging. 2020;68:124-30.
- [CrossRef] [PubMed] [Google Scholar]
- A deep learning image reconstruction algorithm for improving image quality and hepatic lesion detectability in abdominal dual-energy computed tomography: Preliminary results. J Digit Imaging. 2023;36:2347-55.
- [CrossRef] [PubMed] [Google Scholar]
- Improving image quality with deep learning image reconstruction in double-low-dose head CT angiography compared with standard dose and adaptive statistical iterative reconstruction. Br J Radiol. 2023;96:20220625.
- [CrossRef] [PubMed] [Google Scholar]
- Deep learning image reconstruction for improving image quality of contrast-enhanced dual-energy CT in abdomen. Eur Radiol. 2022;32:5499-507.
- [CrossRef] [PubMed] [Google Scholar]
- Application of deep learning image reconstruction-high algorithm in one-stop coronary and carotid-cerebrovascular CT angiography with low radiation and contrast doses. Quant Imaging Med Surg. 2024;14:1860-72.
- [CrossRef] [PubMed] [Google Scholar]
- Deep learning image reconstruction algorithm for carotid dual-energy computed tomography angiography: Evaluation of image quality and diagnostic performance. Insights Imaging. 2022;13:182.
- [CrossRef] [PubMed] [Google Scholar]
- Deep learning reconstruction algorithm and high-concentration contrast medium: Feasibility of a double-low protocol in coronary computed tomography angiography. Eur Radiol. 2025;35:2213-21.
- [CrossRef] [PubMed] [Google Scholar]
- State-of-the-art CT and MR imaging and assessment of atherosclerotic carotid artery disease: The reporting-a consensus document by the European society of cardiovascular radiology (ESCR) Eur Radiol. 2023;33:1088-101.
- [CrossRef] [PubMed] [Google Scholar]
- Global, regional, and national progress towards the 2030 global nutrition targets and forecasts to 2050: A systematic analysis for the Global burden of disease study 2021. Lancet. 2025;404:2543-83.
- [Google Scholar]
- The application value of a vendor-specific deep learning image reconstruction algorithm in “triple low” head and neck computed tomography angiography. Quant Imag Med Surg. 2024;14:2955-67.
- [CrossRef] [PubMed] [Google Scholar]