Skip to main content
Download PDF

Reduced-dose deep learning iterative reconstruction for abdominal computed tomography with low tube voltage and tube current

BMC Medical Informatics and Decision Making volume 24, Article number: 389 (2024) Cite this article

Abstract

Background

The low tube-voltage technique (e.g., 80 kV) can efficiently reduce the radiation dose and increase the contrast enhancement of vascular and parenchymal structures in abdominal CT. However, a high tube current is always required in this setting and limits the dose reduction potential. This study investigated the feasibility of a deep learning iterative reconstruction algorithm (Deep IR) in reducing the radiation dose while improving the image quality for abdominal computed tomography (CT) with low tube voltage and current.

Methods

Sixty patients (male/female, 36/24; Age, 57.72 ± 10.19 years) undergoing the abdominal portal venous phase CT were randomly divided into groups A (100 kV, automatic exposure control [AEC] with reference tube-current of 213 mAs) and B (80 kV, AEC with reference of 130 mAs). Images were reconstructed via hybrid iterative reconstruction (HIR) and Deep IR (levels 1–5). The mean CT and standard deviation (SD) values of four regions of interest (ROI), i.e. liver, spleen, main portal vein and erector spinae at the porta hepatis level in each image serial were measured, and the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were calculated. The image quality was subjectively scored by two radiologists using a 5-point criterion.

Results

A significant reduction in the radiation dose of 69.94% (5.09 ± 0.91 mSv vs. 1.53 ± 0.37 mSv) was detected in Group B compared with Group A. After application of the Deep IR, there was no significant change in the CT value, but the SD gradually increased. Group B had higher CT values than group A, and the portal vein CT values significantly differed between the groups (P < 0.003). The SNR and CNR in Group B with Deep IR at levels 1–5 were greater than those in Group A and significantly differed when HIR and Deep IR were applied at levels 1–3 of HIR and Deep IR (P < 0.003). The subjective scores (distortion, clarity of the portal vein, visibility of small structures and overall image quality) with Deep IR at levels 4–5 in Group B were significantly higher than those in group A with HIR (P < 0.003).

Conclusion

Deep IR algorithm can meet the clinical requirements and reduce the radiation dose by 69.94% in portal venous phase abdominal CT with a low tube voltage of 80 kV and a low tube current. Deep IR at levels 4–5 can significantly improve the image quality of the abdominal parenchymal organs and the clarity of the portal vein.

Peer Review reports

Background

Multiple-phase contrast-enhanced abdominal CT is an indispensable diagnostic tool for evaluating various abdominal diseases [1, 2]. However, concerns remain about the potential risks of ionizing radiation, particularly for populations that may require serial CT examinations [3,4,5]. Previous studies have reported that the low tube-voltage technique (e.g., 80 kV) can efficiently reduce the radiation dose and increase the contrast enhancement of vascular and parenchymal structures in abdominal contrast-enhanced CT [6]. However, in this setting, a high tube current is always required to reduce the image noise and increase the susceptibility to beam hardening artifacts that occur by reducing the photon flux and energy [7,8,9]. This limits the further dose reduction at low tube voltages concurrently with low tube currents during abdominal CT with internally low contrast in soft tissues.

In recent years, model-based iterative reconstruction (MBIR) and deep learning reconstruction algorithms have been demonstrated to play vital roles in denoising the low-dose CT images [10, 11]. In principle, manual and empirical value determined regularization functions are required for MBIR to reduce the image noise and artifacts. It is these facts that may change the image texture, and result in “blurred”, “plastic”, or “cartoonish” artifacts [12]. Notably, most deep learning reconstruction algorithms use high-dose FBP or MBIR images as the target data to replicate their noise textures and visual impressions [13, 14], however, low-contrast or high-density anatomical structures such as the abdomen and pelvis are easily overwhelmed by noise in ultra-low-dose images, which affects the clinical diagnosis. To overcome the above limitations, a deep learning iterative reconstruction (Deep IR) that incorporates both MBIR and deep neural convolutional networks was developed to suppress the noise and preserve texture, and has shown remarkable performance in improving the image quality of low-dose CT images [15,16,17]. In particular, chest and abdomen CT has drawn a great deal of clinical interests, with numerous studies confirming that the Deep IR algorithm can reduce the image noise and improve the image quality of the reduced-dose chest CT and CT angiography [18, 19].

Therefore, we hypothesize that the radiation dose can be efficiently reduced without degrading the image quality by using a low tube voltage and Deep IR algorithm for enhanced CT of the abdomen. The purpose of this study is to explore the feasibility of 80 kV combined with the Deep IR algorithm to reduce the radiation dose and improve the image quality of low-dose abdomen enhanced CT, compared with the commonly used HIR algorithm.

Methods

Patient population

Patients who underwent contrast-enhanced abdominal CT at our institution from March 2023 to June 2023 were prospective enrolled in the study. The inclusion criteria were as follows: (1) patients with clinical need for upper abdominal enhanced CT examination and (2) patients without a history of severe contrast media allergy or renal insufficiency. The exclusion criteria were as follows: (1) patients whose image quality was affected by poor respiratory coordination; (2) patients with metal implants in the abdomen; and (3) patients age < 18 years. Two patients’ images affected by poor respiratory coordination and a patient’s images with substantial metal artifacts at the fundus of stomach were excluded. Ultimately, 60 patients were included in the study and were randomly divided into two groups, labeled A and B. This study was approved by the Hospital Ethics Committee, and all patients were informed of the examination precautions and signed informed consent forms.

Imaging technique and postprocessing

All patients underwent contrast-enhanced abdominal CT on a 320-detector row CT scanner (uCT 960+, United Imaging Healthcare, Shanghai, China). Patients underwent breath-hold training prior to the scan to reduce respiratory motion artifacts. Patients were also informed of the examination precautions. Group A adopted the routine scanning protocol of our institution, and the protocol was as follows: tube voltage 100 kV, AEC, and tube current setting level 3 (reference tube current: 213 mAs). The low-dose scanning scheme for Group B during the portal venous phase was as follows: tube voltage 80 kV, automatic exposure control (AEC), and tube current setting level 2 (reference tube current: 130 mAs); the arterial and delayed phase scanning schemes were the same as those Group A. The pitch and rotation speeds for the two groups were the same, at 0.9937 and 0.5 s/r, respectively. The scope of the scan was from the upper extent of the diaphragm to the upper edge of the pelvis. Contrast agent of 370 mgI/ml of (Ultravist, Bayer, Healthcare Ltd, Guangzhou, China) was used. The volume was selected according to the participants’ body weight at 1.2 ml/kg, with an allowable dose range of 50–95 ml and an injection speed of 2.7 ml/s. The abdominal aorta was monitored via the bolus tracking technique with a threshold of 150 HU and artery, portal and delayed phase scanning was commenced at 16 s, 50 s and 120 s, respectively, after threshold triggering. After scanning, the raw data from the portal phase were transferred to the Explorer Platform (uInnovation-CT, R001, United Imaging Healthcare, Shanghai, China) for reconstruction and analysis. In group A, the hybrid iterative reconstruction (HIR) algorithm [20] was used at a strength level 5 (with the maximum being 10) for reconstruction, and in group B, the Deep IR algorithm was applied at levels 1 to 5 for reconstruction. In both groups, the slice thickness was 1 mm, and the slice interval was 1 mm.

Deep learning iterative reconstruction

Deep IR is the latest-generation deep-learning based algorithm utilizing the newly designed backbone for CT image reconstruction. Technically, Deep IR combines the advantages of deep-learning methods and MBIR, where the regularization term in MBIR is replaced by a convolutional neural network. Therefore, this algorithm not only allows for a substantial noise suppression without inducing plastic image appearance that is easily caused by the regularization term, but also retains the ability of characterizing image detail provided by MBIR. In addition, Deep IR provides five reconstruction strength levels to control the amount of noise reduction, which could be adjusted by radiologists according to their preference and the scanning body parts.

Radiation dose

The volume CT dose index (CTDIvol) and dose–length product (DLP) were recorded, and the effective dose (ED) was calculated for the two groups using the formula ED = DLP × k, where k represents the radiation dose conversion factor, which equals 0.015 mSv/(mGy.cm) for the abdomen.

Image quality evaluation

Objective image quality assessment

The CT values and standard deviations (SDs) of the liver parenchyma, spleen, main portal vein, and left paravertebral muscle were measured by a radiologist (with 5 years of experience in CT imaging). The region of interest (ROI) avoided interfering factors that might affect the measurement results, such as vessels, lesions, and obvious local fat deposition, to ensure that the shape, size, and location of each ROI were basically consistent in each part and under the reconstruction method. The ROI size was 30–50 mm2. All of the data were measured three times, and the average value was taken. The signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of the liver parenchyma, spleen and portal vein were calculated. As the muscle SD value was relatively constant and not strongly affected by contrast enhancement, the muscle SD was selected to represent the image noise.

$$\:\text{SNR=}\frac{{\text{CTvalue}}_{\text{target}}}{{\text{SD}}_{\text{target}}}\text{}$$
$$\:\text{CNR=}\frac{{\text{CTvalue}}_{\text{target}}\text{-}{\text{CTvalue}}_{\text{muscle}}}{{\text{SD}}_{\text{muscle}}}$$

Subjective image quality assessment

The six groups of reconstructed images were independently evaluated by two qualified radiologists (with 5 years and 10 years of experience in CT imaging) with no knowledge of the method of image reconstruction. A five-point scoring method was used to evaluate each of five characteristics: image noise, distortion, clarity of the portal vein, visibility of small structures and overall image quality [21, 22]. The specific scoring criteria for the five areas are shown in Table 1. The final image quality scores were obtained by consensus between the two radiologists if there were discrepancies in their scores through joint reading. An overall image quality of 3 points or above was considered to meet the needs of clinical diagnosis.

Table 1 The subjective score criteria of image quality
Full size table

Statistical analysis

SPSS statistical software (Windows v.22.0, SPSS, Chicago, Illinois) was used for statistical analysis. All of the measured values are expressed as the mean ± SD. The Kolmogorov-Smirnov test was used to test the normality of continuous data. The differences in general data (except for sex, which was tested using the χ2 test), tube current and radiation dose between group A and group B were compared using the independent-samples t test. The objective parameters, including CT values, SD values, SNRs, and CNRs, were analyzed via using one-way ANOVA. The subjective scores were compared using the Kruskal‒Wallis H test. P < 0.05 was considered statistically significant. The Bonferroni correction was used for post hoc pairwise comparisons, and the significance level was adjusted to P < 0.05/15. The consistency of the subjective scores of the two radiologists was analyzed using the kappa test, with kappa ≥ 0.75 representing good consistency, 0.4 ≤ kappa < 0.75 representing general consistency, and kappa < 0.4 representing poor consistency between the two reviewers.

Results

General clinical data and radiation dose

There was no statistically significant difference between the two groups in any of the general variables, including sex, age, height, weight, and BMI (P > 0.05). Differences in mAs, CTDIvol, DLP, and ED between Group A and Group B were statistically significant (P < 0.05). The ED (1.53 ± 0.37 mSv) in Group B was lower than that in Group A (5.09 ± 0.91 mSv) by 69.94%. The specific clinical data and radiation dose parameters are shown in Table 2.

Table 2 Comparison of general data, scan parameters, and radiation dose measurements between the two groups
Full size table

Objective image quality

As the Deep IR level increased from 1 to 5, the portal-phase CT values of the liver, spleen, portal vein and muscle in group B remained basically the same, whereas the SD value of the muscle gradually increased. The SNRs and CNRs of the liver, spleen and portal vein decreased as the reconstruction grade increased. The CT values of the liver, spleen, portal vein and muscle were numerically greater in Group B than in Group A, and the difference in the CT value of the portal vein was statistically significant (P < 0.003). The SNR and CNR in Group B were greater than those in Group A, and the differences between HIR and Deep IR at levels 1 to 3 were significant (P < 0.003). The results are shown in detail in Table 3.

Table 3 Comparison of CT, SNR and CNR values between the HIR and deep IR algorithms
Full size table

Subjective image quality

The intergroup differences in the five subjectively evaluated measures of the six reconstructed images were statistically significant (P < 0.001), and the results are shown in Table 4, and Fig. 1. The image noise scores of Group B were higher than those of group A. The lower the level of Deep IR was, the lower the image noise. The subjective scores of image noise in Group B were higher than those in Group A. The scores of Deep IR levels 4 and 5 in Group B were significantly better than those in Group A in terms of distortion, clarity of the portal vein, visibility of small structures and overall image quality (P < 0.003). Deep IR at levels 1 and 2 increased the distortion and reduced the overall image quality (P < 0.003). (Figures 2, 3 and 4). There was good agreement between the two reviewers (the kappa value was between 0.73 and 0.91).

Table 4 Comparison of subjective scores between HIR and different levels of deep IR
Full size table
Fig. 1
figure 1

Subjective image quality score

Full size image
Fig. 2
figure 2

a Image from a 64-year-old man reconstructed with HIR. The vessel and lesion are moderately clear and sharp, and the noise and overall image quality were rated 3 and 4, respectively. b-f Images from a 65-year-old man reconstructed with Deep IR at levels 5 (b), 4 (c), 3 (d), 2 (e) and 1 (f). The vessel and lesion were very clear and sharp; the noise of the images b-f was rated as 4, 4, 4, 5 and 5, respectively; and the overall image quality of images b-f was rated as 5, 5, 4, 4 and 3, respectively

Full size image
Fig. 3
figure 3

a Image from a 60-year-old woman reconstructed with HIR. The metal boundary was minor distortion, and rated the distortion as 4. b-f Images from a 65-year-old man reconstructed with Deep IR at levels 5 (b), 4 (c), 3 (d), 2 (e) and 1 (f). The metal boundary was not distorted, and the distortion b-f was rated as 5, 5, 5, 4 and 3, respectively

Full size image
Fig. 4
figure 4

a Volume render (VR) image from a 64-year-old man reconstructed with HIR. The branches of the portal vein are moderately clear, and the clarity of the portal vein was rated 3. b-f VR images from a 42-year-old man reconstructed with Deep IR at levels 5 (b), 4 (c), 3 (d), 2 (e) and 1 (f). The branches of the portal vein are very clear in the images, and the clarity of the portal vein in images b-f was rated 5, 5, 5, 5 and 4, respectively

Full size image

Discussion

Dose optimization is highly important for eliminating the possible adverse effects of ionizing radiation exposure from CT. In this study, we employed the Deep IR algorithm to further reduce the radiation dose by using the combined low tube-voltage and tube-current technique. Our results indicated that our protocol with Deep IR algorithm could significantly reduce the radiation dose (5.09 ± 0.91 mSv vs. 1.53 ± 0.37 mSv) in portal-phase abdominal CT while improving the image quality. Specifically, Deep IR at levels 4 and 5 enabled a 69.69% reduction in the radiation dose while still providing improved SNR, CNR, and subjectively scored image quality. It demonstrates the potential of deep learning techniques to improve patient safety and diagnostic utility in abdominal CT evaluation.

The radiation dose is known to be proportional to the square of the tube voltage [23]. A previous showed that [24] when the tube current is constant and the tube voltage is reduced from 120 kV to 100 kV, the radiation dose can be reduced by 33%; The radiation dose can be reduced by 65% when the tube voltage is reduced from 120 kV to 80 kV, and the radiation dose can be effectively reduced by reducing the tube voltage. Seung et al. [25] showed that the use of 80 kV could reduce the radiation dose by 45.2% with comparable or improved image quality in patients with abdominal tumors. Li et al. [26] used 80 kV scans combined with deep learning image reconstruction (DLIR) for late arterial-phase abdominal CT of patients with a low BMI, and the results showed that DLIR could significantly reduce the image noise and improve the overall image quality of low-dose abdominal CT while reducing the radiation dose by 57%. Many studies [27, 28] have shown that 100 kV combined with HIR can obtain high-quality abdominal images. 100 kV and HIR are routinely used for scanning in our institution. In Group B, not only did the tube voltage decrease from 100 kV to 80KV, but the tube current also decreased from 205 mAs to 154 mAs. The results showed that the combination of reducing the tube voltage and tube current with the image reconstruction algorithm can better reduce the radiation dose (by 69.94%) while ensuring the same or better abdominal image quality.

X-ray energy decreases and approaches the K edge of iodine (33.2 keV). Through the photoelectric effect, X-ray energy interacts more with iodine-based materials, which can cause greater X-ray attenuation and greater contrast enhancement. Yu et al. [29] conducted a phantom experiment and showed that the iodine attenuation value of 80 kV CT increased by 70% and 100%, on average, compared with that of 120 kV CT and 140 kV CT. Zamboni et al. [30] found that pancreatic cancer was more obvious when a low kV was used. The results of this study showed that the CT values of all organs in Group B were greater than those in Group A, and the CT values of the portal vein were significantly greater than those in Group A, which is 33 HU greater that those in Group A. Sufficient CT values can increase the contrast of images and optimize the display of vessels. In addition, patients in both groups were treated with high-concentration iodine contrast agent (370 mgI/ml). Iezzi et al. [31] found that the use of a high-concentration contrast agent and low kV can maintain the image quality of the abdominal aorta and reduce the radiation dose by 74%. A low tube voltage combined with a high concentration of iodine contrast agent can not only reduce the radiation dose, but also improve the image quality of solid organs and ensure the SNR. In this study, the SNRs and CNRs of the liver, spleen and portal vein in Group B were greater than those in Group A, which effectively enhanced the display of parenchymal organs, vessels and their branches and improved the contrast between the portal vein and its surrounding tissues.

A decrease in tube voltage and tube current produces remarkable image noise and is sensitive to hardening artifacts, especially in abdominal CT with low soft tissue contrast, and an increase in image noise leads to a decrease in image quality, thus affecting the diagnostic accuracy [6]. The Deep IR algorithm can automatically identify useful signals and noise in images, which can not only remove fringe artifacts in low-dose images but also achieve image denoising so that the reconstructed image has better contrast and detail display [32]. In this study, compared with the HIR algorithm, the Deep IR algorithm significantly improved the SNRs and CNRs of abdominal organs and the portal vein at all reconstruction levels. The lower the Deep IR reconstruction level was, the higher the SNR and CNR. However, there was a discrepancy between objective image quality and subjective scoring results. Although higher SNR and CNR were obtained on Deep IR images with lower reconstruction strengths, readers more preferred those images with higher strengths. This was because Deep IR with low strengths provide lower image noise compared to that with high strengths; thereby obtaining better objective image quality. Nevertheless, it should be noted that such objective metrics, such as SNR and CNR are simple metrics related with image noise, which may not be sufficient for assessing image quality thoroughly [22, 33, 34]. In this study, when the Deep IR algorithm was applied at strength levels 1 and 2, although the image noise was significantly reduced, the reconstructed images were subject to distortion and other alterations due to blurring effects and texture changes, resulting in lower subjective scoring. On the basis of subjective and objective evaluation, appropriate Deep IR reconstruction levels can effectively improve the image quality of abdominal parenchymal organs and the clarity of the portal vein, and Deep IR grades 4 and 5 have the best performance.

The results of this study have significant clinical implications, and the combined use of low tube voltage, low tube current and Deep IR techniques may improve patient safety by reducing radiation exposure while improving image quality of abdominal CT and angiography. This is particularly beneficial for patients who require regular follow-up, those who require multi-phase, large-scale perfusion scans, and those who are sensitive to radiation. This study also provides a reference for the subsequent research on enhanced abdominal CT scanning with lower doses and more advanced reconstruction algorithms.

Some limitations still exist in this study. First, this was a single-center study with a relatively small sample size. Second, the maximum BMI of the patients in this study was 32, and the feasibility of improving abdominal image quality in patients with high BMIs needs to be further studied. Finally, only the image quality of the portal venous phase of the abdomen was studied, and the sensitivity and specificity of lesion detection and differentiation were not evaluated in this study. The next step will be extended to the arterial phase and delayed phase of the abdomen to achieve the goal of multiphase enhanced CT with a low radiation dose.

Conclusion

The Deep IR algorithm can meet the clinical requirements and reduce the radiation dose by 69.94% in portal venous phase abdominal CT with a low tube voltage of 80 kV and a low tube current. Deep IR at levels 4–5 can significantly improve the image quality of the abdominal parenchymal organs and the clarity of the portal vein.

Data availability

Authors declares that all data and material will be available for further research and verification by contacting corresponding author.

Abbreviations

Deep IR:

Deep Learning Iterative Reconstruction Algorithm

CT:

Computer Tomography

AEC:

Automatic Exposure Control

HIR:

Hybrid Iterative Reconstruction

SD:

Standard Deviation

ROI:

Regions of Interest

SNR:

Signal-to-Noise Ratio

CNR:

Contrast to Noise Ratio

MBIR:

Model-Based Iterative Algorithm

CTDIvol:

Volume CT Dose Index

DLP:

Dose Length Product

ED:

Effective Dose

DLIR:

Deep Learning Image Reconstruction

References

  1. Jensen CT, Wagner-Bartak NA, Vu LN, Liu X, Raval B, Martinez D, et al. Detection of colorectal hepatic metastases is superior at standard radiation dose CT versus reduced dose CT. Radiology. 2019;290(2):400–9.

  2. Park S, Yoon JH, Joo I, Yu MH, Kim JH, Park J, et al. Image quality in liver CT: low-dose deep learning vs standard-dose model-based iterative reconstructions. Eur Radiol. 2022;32(5):2865–74.

    Article  CAS  PubMed  Google Scholar 

  3. Hong JY, Han K, Jung JH, Kim JS. Association of exposure to Diagnostic Low-Dose Ionizing Radiation with Risk of Cancer among youths in South Korea. JAMA NETW OPEN. 2019;2(9):e1910584.

    Article  PubMed  PubMed Central  Google Scholar 

  4. Sakane H, Ishida M, Shi L, Fukumoto W, Sakai C, Miyata Y, et al. Biological effects of low-dose chest CT on chromosomal DNA. Radiology. 2020;295(2):439–45.

    Article  PubMed  Google Scholar 

  5. Rogalla P, Paravasthu M, Farrell C, Kandel S. Helical CT with variable target noise levels for dose reduction in chest, abdomen and pelvis CT. Eur Radiol. 2018;28(9):3922–8.

    Article  PubMed  Google Scholar 

  6. Marin D, Nelson RC, Schindera ST, Richard S, Youngblood RS, Yoshizumi TT, et al. Low-tube-voltage, high-tube-current multidetector abdominal CT: improved image quality and decreased radiation dose with adaptive statistical iterative reconstruction algorithm–initial clinical experience. Radiology. 2010;254(1):145–53.

    Article  PubMed  Google Scholar 

  7. Yin X, Zhao Q, Liu J, Yang W, Yang J, Quan G, et al. Domain progressive 3D residual convolution network to improve low-dose CT imaging. IEEE Trans Med Imaging. 2019;38(12):2903–13.

    Article  Google Scholar 

  8. Xu Y, Zhang TT, Hu ZH, Li J, Hou HJ, Xu ZS, et al. Effect of iterative reconstruction techniques on image quality in low radiation dose chest CT: a phantom study. Diagn Interv Radiol. 2019;25(6):442–50.

    Article  PubMed  PubMed Central  Google Scholar 

  9. Wang X, Zheng F, Xiao R, Liu Z, Li Y, Li J, Zhang X, Hao X, Zhang X, Guo J, et al. Comparison of image quality and lesion diagnosis in abdominopelvic unenhanced CT between reduced-dose CT using deep learning post-processing and standard-dose CT using iterative reconstruction: a prospective study. EUR J RADIOL. 2021;139:109735.

    Article  PubMed  Google Scholar 

  10. Zhong J, Shen H, Chen Y, Xia Y, Shi X, Lu W, et al. Evaluation of image quality and detectability of deep learning image reconstruction (DLIR) algorithm in single- and dual-energy CT. J Digit Imaging. 2023;36(4):1390–1407.

    Article  PubMed  PubMed Central  Google Scholar 

  11. Nagata M, Ichikawa Y, Domae K, Yoshikawa K, Kanii Y, Yamazaki A, et al. Application of deep learning-based denoising technique for radiation dose reduction in dynamic abdominal CT: comparison with standard-dose CT using hybrid iterative reconstruction method. J Digit Imaging. 2023;36(4):1578–87.

    Article  PubMed  PubMed Central  Google Scholar 

  12. Geyer LL, Schoepf UJ, Meinel FG, Nance JJ, Bastarrika G, Leipsic JA, et al. State of the art: iterative CT reconstruction techniques. Radiology. 2015;276(2):339–57.

    Article  PubMed  Google Scholar 

  13. Zhang Y, Yu H. Convolutional neural network based metal artifact reduction in X-Ray computed Tomography. IEEE Trans Med Imaging. 2018;37(6):1370–81.

    Article  Google Scholar 

  14. Jensen CT, Gupta S, Saleh MM, Liu X, Wong VK, Salem U, et al. Reduced-dose deep learning reconstruction for abdominal CT of liver metastases. Radiology. 2022;303(1):90–8.

    Article  PubMed  Google Scholar 

  15. Yang L, Liu H, Han J, Xu S, Zhang G, Wang Q, et al. Ultra-low-dose CT lung screening with artificial intelligence iterative reconstruction: evaluation via automatic nodule-detection software. Clin Radiol. 2023;78(7):525–31.

    Article  CAS  PubMed  Google Scholar 

  16. Gong H, Peng L, Du X, An J, Peng R, Guo R, et al. Artificial intelligence iterative reconstruction in computed tomography angiography: an evaluation on pulmonary arteries and aorta with routine dose settings. J Comput Assist Tomo. 2024;48(2):244–50.

  17. Zeng L, Xu X, Zeng W, Peng W, Zhang J, Sixian H, Liu K, Xia C, Li Z. Deep learning trained algorithm maintains the quality of half-dose contrast-enhanced liver computed tomography images: comparison with hybrid iterative reconstruction: study for the application of deep learning noise reduction technology in low dose. EUR J RADIOL. 2021;135:109487.

    Article  PubMed  Google Scholar 

  18. Wang Q, Xu S, Zhang G, Zhang X, Gu J, Yang S, Zeng M, Zhang Z. Applying a CT texture analysis model trained with deep-learning reconstruction images to iterative reconstruction images in pulmonary nodule diagnosis. J APPL CLIN MED PHYS. 2022;23(11):e13759.

    Article  PubMed  PubMed Central  Google Scholar 

  19. Li W, You Y, Zhong S, Shuai T, Liao K, Yu J, Zhao J, Li Z, Lu C. Image quality assessment of artificial intelligence iterative reconstruction for low dose aortic CTA: a feasibility study of 70 kVp and reduced contrast medium volume. EUR J RADIOL. 2022;149:110221.

    Article  PubMed  Google Scholar 

  20. Li J, Wang X, Huang X, Chen F, Zhang X, Liu Y, Luo G, Xu X. Application of CareDose 4D combined with Karl 3D technology in the low dose computed tomography for the follow-up of COVID-19. BMC MED IMAGING. 2020;20(1):56.

    Article  PubMed  PubMed Central  Google Scholar 

  21. Lyu P, Liu N, Harrawood B, Solomon J, Wang H, Chen Y, et al. Is it possible to use low-dose deep learning reconstruction for the detection of liver metastases on CT routinely? Eur Radiol. 2023;33(3):1629–40.

    Article  CAS  PubMed  Google Scholar 

  22. Ren Z, Zhang X, Hu Z, Li D, Liu Z, Wei D, et al. Reducing radiation dose and improving image quality in CT portal venography using 80 kV and adaptive statistical iterative reconstruction-V in slender patients. Acad Radiol. 2020;27(2):233–43.

    Article  PubMed  Google Scholar 

  23. Lee HN, Lee SM, Choe J, Lee SM, Chae EJ, Do KH, et al. Diagnostic performance of CT-guided percutaneous transthoracic core needle biopsy using low tube voltage (100 kVp): comparison with conventional tube voltage (120 kVp). ACTA RADIOL. 2018;59(4):425–33.

    Article  PubMed  Google Scholar 

  24. Seyal AR, Arslanoglu A, Abboud SF, Sahin A, Horowitz JM, Yaghmai V. CT of the abdomen with reduced tube voltage in adults: a practical approach. Radiographics. 2015;35(7):1922–39.

    Article  PubMed  Google Scholar 

  25. Choi SJ, Ahn SJ, Park SH, Park SH, Pak SY, Choi JW, Shim YS, Jeong YM, Kim B. Dual-source abdominopelvic computed tomography: comparison of image quality and radiation dose of 80 kVp and 80/150 kVp with tin filter. PLoS ONE. 2020;15(9):e231431.

    Article  Google Scholar 

  26. Li LL, Wang H, Song J, Shang J, Zhao XY, Liu B. A feasibility study of realizing low-dose abdominal CT using deep learning image reconstruction algorithm. J X-ray Sci Technol. 2021;29(2):361–72.

    Google Scholar 

  27. Cheng Y, Han Y, Li J, Fan G, Cao L, Li J, Jia X, Yang J, Guo J. Low-dose CT urography using deep learning image reconstruction: a prospective study for comparison with conventional CT urography. BRIT J RADIOL. 2021;94(1120):20201291.

    Article  PubMed  PubMed Central  Google Scholar 

  28. Ippolito D, Maino C, Pecorelli A, Salemi I, Gandola D, Riva L, Talei FC, Sironi S. Application of low-dose CT combined with model-based iterative reconstruction algorithm in oncologic patients during follow-up: dose reduction and image quality. BRIT J RADIOL. 2021;94(1124):20201223.

    Article  PubMed  PubMed Central  Google Scholar 

  29. Yu L, Bruesewitz MR, Thomas KB, Fletcher JG, Kofler JM, McCollough CH. Optimal tube potential for radiation dose reduction in pediatric CT: principles, clinical implementations, and pitfalls. Radiographics. 2011;31(3):835–48.

  30. Zamboni GA, Ambrosetti MC, Guariglia S, Cavedon C, Pozzi MR. Single-energy low-voltage arterial phase MDCT scanning increases conspicuity of adenocarcinoma of the pancreas. Eur J Radiol. 2014;83(3):e113–7.

    Article  PubMed  Google Scholar 

  31. Iezzi R, Santoro M, Marano R, Di Stasi C, Dattesi R, Kirchin M, et al. Low-dose multidetector CT angiography in the evaluation of infrarenal aorta and peripheral arterial occlusive disease. Radiology. 2012;263(1):287–98.

    Article  PubMed  Google Scholar 

  32. Liu J, Zhang Y, Zhao Q, Lv T, Wu W, Cai N, et al. Deep iterative reconstruction estimation (DIRE): approximate iterative reconstruction estimation for low dose CT imaging. Phys Med Biol. 2019;64(13):135007.

    Article  PubMed  Google Scholar 

  33. Yoshida K, Nagayama Y, Funama Y, Ishiuchi S, Motohara T, Masuda T, et al. Low tube voltage and deep-learning reconstruction for reducing radiation and contrast medium doses in thin-slice abdominal CT: a prospective clinical trial. Eur Radiol. 2024;34(11):7386–96.

  34. Mileto A, Guimaraes LS, McCollough CH, Fletcher JG, Yu L. State of the art in abdominal CT: the limits of iterative reconstruction algorithms. Radiology. 2019;293(3):491–503.

Download references

Acknowledgements

Not applicable.

Funding

This work was supported by National Natural Science Foundation of China (Grant Number: 12226007).

Author information

Authors and Affiliations

  1. Department of Radiology, The First Affiliated Hospital of Xi’an Jiaotong University, Xi’an, 710061, P. R. China

    Shumeng Zhu, Baoping Zhang, Qian Tian, Ao Li, Zhe Liu, Wei Hou, Wenzhe Zhao, Xin Huang, Yao Xiao, Yiming Wang, Rui Wang, Yuhang Li, Jian Yang & Chao Jin

  2. Shaanxi Engineering Research Center of Computational Imaging and Medical Intelligence, Xi’an, 710061, P. R. China

    Shumeng Zhu, Baoping Zhang, Qian Tian, Ao Li, Zhe Liu, Wei Hou, Wenzhe Zhao, Xin Huang, Yao Xiao, Yiming Wang, Rui Wang, Yuhang Li, Jian Yang & Chao Jin

Authors
  1. Shumeng Zhu
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Baoping Zhang
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Qian Tian
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. Ao Li
    View author publications

    You can also search for this author inPubMed Google Scholar

  5. Zhe Liu
    View author publications

    You can also search for this author inPubMed Google Scholar

  6. Wei Hou
    View author publications

    You can also search for this author inPubMed Google Scholar

  7. Wenzhe Zhao
    View author publications

    You can also search for this author inPubMed Google Scholar

  8. Xin Huang
    View author publications

    You can also search for this author inPubMed Google Scholar

  9. Yao Xiao
    View author publications

    You can also search for this author inPubMed Google Scholar

  10. Yiming Wang
    View author publications

    You can also search for this author inPubMed Google Scholar

  11. Rui Wang
    View author publications

    You can also search for this author inPubMed Google Scholar

  12. Yuhang Li
    View author publications

    You can also search for this author inPubMed Google Scholar

  13. Jian Yang
    View author publications

    You can also search for this author inPubMed Google Scholar

  14. Chao Jin
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

SZ and CJ wrote the main manuscript text. JY, BZ, ZL, and YL contributed to the conception of the study. QT, AL, and WH performed the data analyses. JY, WZ, and XH helped perform the analysis with constructive discussions. YX, YW, and RW prepared the figures and tables. All of the authors read and approved the final manuscript.

Corresponding authors

Correspondence to Jian Yang or Chao Jin.

Ethics declarations

Ethics approval and consent to participate

The current study was approved by a Research Ethics Committee of the First Affiliated Hospital of Xi ‘an Jiaotong University [Code: XJTU1AF2021LSB-02]. We declare that all research phases performed in accordance with the Declaration of Helsinki-Ethical Principles for Medical Research Involving Human Subjects. Research participants were informed about the study and informed consent was obtained and the voluntary nature of the participation was explained to participants, and they were assured about anonymity and confidentiality of data collected.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Zhu, S., Zhang, B., Tian, Q. et al. Reduced-dose deep learning iterative reconstruction for abdominal computed tomography with low tube voltage and tube current. BMC Med Inform Decis Mak 24, 389 (2024). https://doi.org/10.1186/s12911-024-02811-w

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12911-024-02811-w

Keywords

BMC Medical Informatics and Decision Making

ISSN: 1472-6947

Contact us