Effects of different combinations of noise index and preset adaptive statistical iterative reconstruction Veo on the accuracy and image quality of bone mineral density measurements using fast kilovolt-switching dual-energy computed tomography: a phantom study

Introduction

Osteoporosis is one of the most common chronic metabolic bone diseases and is characterized by decreased bone mass and impaired bone microarchitecture that leads to increased fragility and fracture susceptibility (1). As the population ages, the severity of osteoporosis and brittle fractures will increase, with the morbidity and mortality increasing without treatment. Bone mineral density (BMD) is an important indicator for the diagnosis osteoporosis (2,3). Therefore, accurate measurement of BMD is vital to the early diagnosis of osteoporosis and reducing the risk of fracture.

Computed tomography (CT) is widely used as a rapid, noninvasive, and cost-effective examination. Quantitative CT (quantitative CT) has high accuracy in the measurement of BMD and is the current gold standard for BMD measurement (4-7). With the increase in juvenile osteoporosis, low-dose measurement of BMD has become an urgent clinical demand (8). Rapid kilovolt (kV)-switching dual-energy CT (DECT) can obtain images reflecting the various base material pairs, and when the base material pairs are exactly the two main components contained in the tissue, the measured base material can indicate the relative content of the substance in the tissue. The value of DECT in measuring vertebral BMD has been widely recognized (9-11), with hydroxyapatite (HAP) (water) being used as the base material pair to measure BMD.

Automatic tube current modulation is a commonly used method for reducing radiation dose in clinical practice. The noise index (NI) is a key index for adjusting tube current output in automatic tube current modulation technology. The NI value represents the image noise size of the scanned area. The setting of the NI for different parts is critical to applying to automatic tube current modulation technology.

Filtered back projection is the most commonly used reconstruction algorithm for CT. However, when low doses are used, high noise in the projection can lead to artifacts and poor contrast. Iterative reconstruction (IR) algorithms can overcome the limitations of filtered back projection, resulting in better image quality at a lower dose. GE HealthCare (Chicago, IL, USA) recently introduced a new IR algorithm involving more advanced noise and target modeling: infrared reconstruction algorithm adaptive statistical iterative reconstruction Veo (ASiR-V).

Huang et al. showed that the material equivalent density image of HAP-based materials generated by DECT to measure the equivalent density in HAP imaging can improve the measurement accuracy of BMD (12). Ye et al. found that with the increase in abdominal adipose content, the accuracy of BMD measurement by DECT scans increased and was superior to that derived by quantitative CT (13). On the basis of these studies, we further adjusted NI and ASiR-V to increase the accuracy and reduce the influence of abdominal adipose content. Studies have reported that DECT scans may be superior to quantitative CT in measuring lumbar BMD (11,14-17). However, BMD may be underestimated under low–radiation dose quantitative CT in the detection of osteoporosis (5,6,11).

NI and ASiR-V are two important scanning parameters for DECT scans (18-22). However, the means to achieving a precise BMD, reducing the radiation dose, and improving the image quality of different body mass index (BMI) groups with DECT scans remains unclear (6,7,12).

Therefore, in this study, we used rapid kV-switching DECT scans to study the accuracy of BMD measurements at different total adipose tissue (TAT) areas under different NI and ASiR-V levels.

Methods

Image acquisition

Imaging data were acquired on a Revolution CT scanner (GE HealthCare) using the European Spine Phantom (ESP; QRM GmbH, Möhrendorf, Germany). The phantom consists of an epoxy resin body with three HAP inserts with densities of 50 mg/cm³ (osteoporotic), 102 mg/cm³ (osteopenia), and 197 mg/cm³ (normal), which were labeled as the first lumbar vertebra (L1), second lumbar vertebra (L2), and third lumbar vertebra (L3), respectively. Three pieces of fresh (within 6 hours after slaughter) porcine-isolated fat (without skin) of different sizes were each selected and wrapped around the phantom (Figure 1) to simulate different levels of human abdominal TAT.

Figure 1 A phantom being scanned with the Revolution CT scanner. CT, computed tomography.

The phantom was scanned by a rapid kV-switching DECT scanner system (Revolution CT, GE Healthcare). The acquisition parameters were as follows: tube current, 200–500 mA (under the automatic tube current modulation technique); NI, ranging from 4 to 20 (at increments of 2); and ASiR-V, ranging from 0% to 100% (20% intervals). The other acquisition parameters were as follows: DECT KV mode with rapid KV switching between 80 and 140 kVp; helical pitch, 0.992; and rotation time, 0.8 s. Each scan was repeated three times. Equivalent material density images using HAP and water as basis materials were reconstructed using the bone kernel with a slice thickness of 1.25 mm.

Data measurement and image evaluation

We evaluated the measurements by examining the effects of abdominal adipose tissue, three-slice BMD quantification, image quantitative assessment, and radiation dose.

The data from rapid kV-switching DECT scans images were postprocessed using an Advanced Workstation 4.6 (GE HealthCare). With HAP (water) used as the base material pair, the base material image of HAP (water) was reconstructed. For all HAP images, regions of interest (ROIs) with a circular area of 15×15 mm² were placed in the middle of L1, L2, and L3, and the averaged value was considered to be the measured BMD (Figure 2). All measurements were performed by two radiologists with at least 3 years of experiences in abdominal imaging. The differences of measured BMD and true BMD are represented by root mean square error (RMSE), which was calculated as follows:

RMSE=∑i=1N(yi−y^i)2N[1]

Figure 2 DECT viewer measurements of the BMD of the first, second, and third lumbar vertebrae. The red circles indicate the axial region of interest, and the yellow squares indicate the sagittal region of interest. DECT, dual-energy computed tomography; BMD, bone mineral density.

For the above reconstructed images, the ROIs were placed in the middle of L1, L2, and L3, and the mean values of CT were recorded as CT₁. The ROIs with the same area were placed on the position two-thirds between the corresponding lumbar vertebrae and anterior abdomen, and corresponding mean values were recorded as CT₂. The standard deviation (SD) of CT₂ was considered to be the background noise of the image. The contrast-to-noise ratio (CNR) was calculated to evaluate the objective image quality, as follows:

CNR=(CT1−CT2)SD×100%[2]

The noise power spectrum is a commonly used performance index to evaluate noise reduction technology in imaging systems. Images reconstructed with and without noise reduction techniques can be compared by their noise power spectrum to better understand the effect of noise reduction techniques on image sound (23,24). Image analysis was performed using imQuest 7.1 software (Duke University, NC, USA). The ROI was placed at the middle layer of the vertebral body. To reduce the influence of random effects, the measurement range was calculated for the continuous layers from the appearance of L1 to the disappearance of L3 vertebrae (including the vertebral body layer) for noise (HU; Hounsfield units). The noise power spectrum was drawn. The noise power spectrum is a spatial resolution function that describes the noise change under Fourier transform and reflects the size of the image noise. The image noise of each group was recorded and its average value was calculated. The CT dose index volume (CTDIvol) of each scan was automatically calculated by computer.

TAT area measurement

On a quantitative CT workstation (Mindways Software Inc., Austin, TX, USA), we used tissue composition analysis to measure the TAT area of the middle level of the L2. The area of the three pieces of porcine-isolated fats was approximately 100, 200, and 350 cm², denoting a slim, normal, and obese abdominal body, respectively (Figure 3).

Figure 3 Quantitative CT measurement of the total adipose tissue area at the second lumbar vertebral center level of the phantom. Quantitative CT coloring of the different components: the green aperture is close to the outer edge of the phantom, the mixed blue-and-yellow part inside the green aperture is the water-equivalent material, the pink part is the vertebrae, and the blue part outside the green aperture is the measured area, which is the total adipose tissue area of the anthropomorphic phantom. CT, computed tomography.

Statistical analysis

All data were analyzed with SPSS 26 (IBM Corp., Armonk, NY, USA). A one-way analysis of variance (ANOVA) was used to compare the BMD under different combinations of NI and ASiR-V levels. A single-tail, one-sample t-test was applied to compare the measured BMD and the known ground truth BMD. Through GraphPad Prism 9.2.0 (GraphPad Software, San Diego, CA, USA) software, an ANOVA Tukey test was conducted to compare the differences of measured values of each vertebral body under different TAT conditions between the two scanning conditions. P<0.05 was considered to indicate statistical significance.

With the TAT group with S=0 cm², the optimal combination of NI =18 and ASiR-V_60% level was identified by comparing the radiation dose, relative error, and image quality. For the default combination, the NI was 6 and the ASiR-V level was 0%. Subsequently, the four TAT groups were scanned under the optimal combination and the default combination.

Results

CTDIvol under different combinations of NI and ASiR-V levels

When 4≤ NI ≤10 and ASiR-V ≥60%, the CTDIvol remained the same, even when the ASiR-V level or NI was changed. For NI ≥12, within this range, CTDIvol decreased with increase in NI and ASiR-V (Figure 4). Overall, increasing NI and ASiR-V to a certain extent resulted in a decrease in radiation dose.

Figure 4 CTDIvol under different combinations of NI and ASiR-V level (when 4≤ NI ≤10, NI =12 ASIR-V_0–40%, NI =14 ASIR-V_0–20%, NI =16–18, and ASiR-V_0%, the CTDIvol was 19.37 mGy; when ASiR-V ≥60%, the CTDIvol was only correlated with NI). CTDIvol, computed tomography dose index volume; NI, noise index; ASiR-V, adaptive statistical iterative reconstruction-Veo; CT, computed tomography.

Comparison of measured BMD under different combinations of NI and ASiR-V levels

There were significant differences in measured BMD values in all vertebrae under different NI and ASiR-V levels. For 54 sets of combinations of NI and ASiR-V in this study, 14 sets demonstrated significant differences between the measured BMD and the true BMD of the L1 (50 mg/cm³); 19 sets demonstrated no significant differences between the two BMDs of the L2 (102 mg/cm³) with 14≤ NI ≤20 in most sets; and 10 sets demonstrated no significant differences between the two BMDs of the L3 (197 mg/cm³) with 16≤ NI ≤20 and ASiR-V ≥40% in most sets. There was no significant difference between the two BMD values of the L1, L2, and L3 values for the following combinations: NI =14 with ASiR-V_80%, NI =16 with ASiR-V_100%, NI =18 with ASiR-V_60%, NI =18 with ASiR-V_80%, NI =20 with ASiR-V_60%, NI =20 with ASiR-V_80%, and NI =20 with ASiR-V_100% (P>0.05) (Table 1).

RMSE of BMD

For all combinations of NI and ASiR-V, the RMSE values of BMD in the L1 and L2 were smaller than 5. With NI ≥14 and ASiR-V ≥20%, the RMSE values of BMD in the L3 were also smaller than 10. With NI =18 and ASiR-V_60%, the RMSE values of BMD in the L1–L3 were all less than 2 (Figure 5).

Figure 5 RMSE of the phantom under different combinations of NI and ASiR-V levels. (A) RMSE of the L1 vertebra. (B) RMSE of the L2 vertebra. (C) RMSE of the L3 vertebra. RMSE, root mean square error; NI, noise index; ASiR-V, adaptive statistical iterative reconstruction Veo.

Image quality evaluation

There was a negative correlation between NI and CNR and a positive correlation between ASiR-V and CNR. Compared to the default combination (NI =6 and ASiR-V_0%), the optimal combination (NI =18 and ASiR-V_60%) yielded superior CNR values (12.27% in L1, 7.84% in L2, and 11.33% in L3). The optimal combination significantly reduced the radiation dose without affecting image quality (Figure 6). When NI =4, the noise with the same ASiR-V weight was the lowest; however, at this time, the CTDIvol was excessively high and did not meet the requirements of low dose examination in clinical practice. When NI ≥8, a positive correlation was observed between NI and noise, while a negative correlation was observed between ASiR-V and noise (Figure 7). The enhancement of image quality could be achieved by reducing NI and increasing ASiR-V.

Figure 6 CNR of the phantom under different combinations of NI and ASiR-V levels. (A) CNR of the L1 vertebra. (B) CNR of the L2 vertebra. (C) CNR of the L3 vertebra. CNR, contrast-to-noise ratio; NI, noise index; ASiR-V, adaptive statistical iterative reconstruction Veo.

Figure 7 Noise of the phantom under different combinations of the NI and ASiR-V levels. NI, noise index; ASiR-V, adaptive statistical iterative reconstruction Veo.

Comparison of the optimal and the default combination under different TAT areas

Under different TAT areas under the default combination (NI =6 and ASiR-V_0%), only the L1 (TAT =0 cm²) showed no significant difference between the measured BMD and the true BMD. Under the optimal combination (NI =18 and ASiR-V_60%), there was no significant difference between the measured BMD and the true BMD in three vertebrae and four TAT areas (Table 2, Figure 8). Both combinations generated a positive correlation between RMSE and TAT area and a negative correlation between the CNR and TAT area. Compared to those under the default combination (NI =6 and ASiR-V_0%), the RMSE and CNR of the optimal combination (NI =18 and ASiR-V_60%) were superior, and the change in these two values was smaller (Figures 9,10). Under the influence of different TAT areas, the noise power spectrum area of the optimal combination (NI =18 and ASiR-V_60%) was smaller compared to the default combination (NI =6 and ASiR-V_0%). As the fat area increased, the default group changed significantly (the noise area increased by 88.3%, from 100 to 350 cm²), while the change in the optimal combination was relatively small (the noise area increased by 22.4% from 100 to 350 cm²) (Figure 11).

Figure 8 Measured BMD of the L1–L3 vertebrae under different TAT conditions using the default combination (NI =6 and ASiR-V_0%) and the optimal combination (NI =18 and ASiR-V_60%). Different BMDs of the (A) L1 vertebra, (B), L2 vertebra, (C) and L3 vertebra. *, P<0.05. BMD, bone mineral density; TAT, total adipose tissue; NI, noise index; ASiR-V, adaptive statistical iterative reconstruction Veo.

Figure 9 RMSE of the phantom in different TAT areas using the default combination (NI =6 and ASiR-V_0%) and the optimal combination (NI =18 and ASiR-V_60%). (A) RMSE of the L1 vertebra. (B) RMSE of the L2 vertebra. (C) RMSE of the L3 vertebra. RMSE, root mean square error; TAT, total adipose tissue; NI, noise index; ASiR-V, adaptive statistical iterative reconstruction Veo.

Figure 10 CNR of the phantom in different TAT areas using the default combination (NI =6 and ASiR-V_0%) and the optimal combination (NI =18 and ASiR-V_60%). (A) CNR of the L1 vertebra. (B) CNR of the L2 vertebra. (C) CNR of the L3 vertebra. CNR, contrast-to-noise ratio; TAT, total adipose tissue; NI, noise index; ASiR-V, adaptive statistical iterative reconstruction Veo.

Figure 11 NPS of the phantom in different TAT areas using the default combination (NI =6 and ASiR-V_0%) and the optimal combination (NI =18 and ASiR-V_60%). NPS, noise power spectrum; TAT, total adipose tissue; NI, noise index; ASiR-V, adaptive statistical iterative reconstruction Veo.

Discussion

For accurate BMD measurements, ensuring image quality meets clinical diagnostic needs and reducing radiation dose to a “reasonably achievable minimum level” have been consistent concerns. Through the adjustment of the ASiR-V weight and NI, the accuracy of BMD by spectral CT can be further ensured while significantly reducing radiation dose without affecting image quality. The aim of our study was to minimize the dose while maintaining image quality by comparing the radiation dose and image quality with different combinations of NI and ASiR-V values.

Accuracy of BMD under different combinations of NI and ASiR-V levels

The phantom used in this study was composed of epoxy resin and HAP, and this composition was equivalent to that of a population with lower BMI. Therefore, a high NI was more suitable, which is consistent with Xiao et al.’s assertion that the NI value should not be set too low in those with a low BMI. Additionally, the RMSE maximum value of the L1–L3 were 0.44, 0.57, and 1.74, respectively, which is within the allowable deviation range (18). The highest accuracy of BMD was derived from the L1 (50 mg/cm³), and was nearly identical to the actual BMD of osteoporosis, meeting the requirement for clinical application. Among the combinations, those with an RMSE less than 2 in the L1–L3 bodies were NI =18 with ASiR-V_60% and NI =18 with ASiR-V_80%.

Radiation dose and image quality under different combinations of NI and ASiR-V levels

In this study, the objective image quality was significantly different (P<0.05) between the different ASiR-V and NI combinations. CNR is directly proportional to ASiR-V and inversely proportional to NI. It makes sense that CNR decreased with increased NI (i.e., CNR was reduced under higher noise), but increased noise reduction should have increased CNR. Studies have shown that a higher ASiR-V could affect the noise texture, with structure details being lost, which is consistent with our findings (19-21). The primary reason for this was that the exposure tube current of each slice was adapted with the scanned body while the automatic tube current modulation was used. Thus, the image noise of each slice was satisfied with the reduction in the radiation dose. However, if ASiR-V ≥80%, waxy artifacts appeared, and the image quality was inferior.

In this study, CTDIvol tended to decrease with the increase in ASiR-V and NI. In some studies, a high NI led to the degradation of image quality and was suggested to be adjusted according to the body shape (24).

We aimed to ensure the accuracy of measured BMD and balance the radiation and image quality. Two parameter groups, NI =18 with ASiR-V_40%, and NI =18 with ASiR-V_60%, were examined in this study, with NI =18 ASiR-V_60% being selected as the best parameter composition. The CTDIvol values of the optimal condition (NI =18 and ASiR-V_60%) and default condition (NI =6 and ASiR-V_0%) were 7.37 mGy and 19.37 mGy, respectively, and a 61.95% reduction in radiation was generated. Compared to that of the default condition, the CNR of the L1–L3 were improved by 12.27%, 7.84%, and 11.33%, respectively.

Effect of TAT on BMD measurements

In previous studies, we examined the effects of different TAT areas on low-dose quantitative CT, and the BMD measured values of each group were significantly different from the true values, suggesting poor clinical practicability (13,25); consequently, we prefer low-dose measurements of BMD with DECT. In this study, we compared the measured BMDs from HAP (water) between the four TAT groups (S=0 cm², S=100 cm², S=200 cm², and S=350 cm²) under the optimal and default conditions. The TAT influenced the accuracy of measured BMD in both conditions. Under the default condition, the RMSE increased successively from the L1 to the L3, and the CNR gradually decreased with the increase in TAT, which is consistent with the findings of Ye et al. (13). The optimal condition demonstrated the same pattern, but the extent of the effect of RMSE and CNR were slighter than those of the default condition. The RMSE of the L1 and L2 was less than 1, while the RMSE of the L3 was less than 6. All CNRs improved by 7–20%, except for the TAT =100 cm² group, in which the CNR only improved by 1–6%.

Under the three TATs of 100 cm², 200 cm², and 350 cm², the noise area of the optimal parameters (NI =18 and ASiR-V_60%) compared to the default parameters (NI =6, ASiR-V_0%) decreased by 4.07%, 30.26%, and 40.74% respectively, indicating that the optimal parameters provided superior scan image quality in obese individuals.

Moreover, under the same TAT condition (except TAT =0 cm²), the RMSE of the L1 was smaller than those of the L2 and L3, indicating that the lower the true BMD, the smaller the RMSE of the measured BMD. Thus, in different BMI groups, rapid kV-switching DECT scans achieved a higher accuracy in the low BMD group, which is also suitable for the diagnosis of osteoporosis.

Certain limitations to this study should be acknowledged. First, this study only involved phantom research, and the potential clinical applications were not examined. Second, the findings were validated in different CT scanners. Third, scanner manufacturers provided dual-source, twin-beam, repeated scans, with dual-layer detector and photon counting for the implementation of DECT, and thus whether the results of rapid kV-switching can be extended to other DECT models remains unclear. Moreover, the average abdominal TAT of the obese population with a vertebral BMD of 80 mg/cm³ is around 350 cm², but the abdominal TAT of some of overweight patients was larger than 350 cm², and these groups should be examined further.

Conclusions

NI =18 with ASiR-V_60% was the optimal combination for measuring the lumbar spine BMD with rapid kV-switching DECT. The accuracy of BMD was the highest, while the radiation dose was reduced and image quality maintained. In clinical practice, precise BMD measurement in different BMI groups can be achieved using low-dose CT by increasing the NI and ASiR-V levels.

Acknowledgments

Funding: None.

Footnote

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-24-185/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Harris K, Zagar CA, Lawrence KV. Osteoporosis: Common Questions and Answers. Am Fam Physician 2023;107:238-46.
2. WHO Scientific Group. World health Organization; 2007. WHO Technical Report; Assessment of Osteoporosis at the Primary Health Care Level. Available online: https://www.sheffield.ac.uk/FRAX/pdfs/WHO_Technical_Report.pdf
3. WHO. Prevention and management of osteoporosis. World Health Organ Tech Rep Ser 2003;921:1-164.
4. Lin W, He C, Xie F, Chen T, Zheng G, Yin H, Chen H, Wang Z. Quantitative CT screening improved lumbar BMD evaluation in older patients compared to dual-energy X-ray absorptiometry. BMC Geriatr 2023;23:231. [Crossref] [PubMed]
5. Brunnquell CL, Winsor C, Aaltonen HL, Telfer S. Sources of error in bone mineral density estimates from quantitative CT. Eur J Radiol 2021;144:110001. [Crossref] [PubMed]
6. Wu Y, Jiang Y, Han X, Wang M, Gao J. Application of low-tube current with iterative model reconstruction on Philips Brilliance iCT Elite FHD in the accuracy of spinal QCT using a European spine phantom. Quant Imaging Med Surg 2018;8:32-8. [Crossref] [PubMed]
7. Zhou S, Chen S, Zhu X, You T, Li P, Shen H, Gao H, He Y, Zhang K. Associations between paraspinal muscles fatty infiltration and lumbar vertebral bone mineral density - An investigation by fast kVp switching dual-energy CT and QCT. Eur J Radiol Open 2022;9:100447. [Crossref] [PubMed]
8. Sakka SD. Osteoporosis in children and young adults. Best Pract Res Clin Rheumatol 2022;36:101776. [Crossref] [PubMed]
9. Kaup M, Wichmann JL, Scholtz JE, Beeres M, Kromen W, Albrecht MH, Lehnert T, Boettcher M, Vogl TJ, Bauer RW. Dual-Energy CT-based Display of Bone Marrow Edema in Osteoporotic Vertebral Compression Fractures: Impact on Diagnostic Accuracy of Radiologists with Varying Levels of Experience in Correlation to MR Imaging. Radiology 2016;280:510-9. [Crossref] [PubMed]
10. Tong X, Fang X, Wang S, Fan Y, Wei W, Xiao Q, Chen A, Liu Y, Liu L. Virtual unenhanced images derived from dual-energy computed tomography for assessing bone mineral density and detecting osteoporosis. Quant Imaging Med Surg 2023;13:6571-82. [Crossref] [PubMed]
11. Wang M, Wu Y, Zhou Y, Dong J, Hou P, Gao J. The new fast kilovoltage-switching dual-energy computed tomography for measuring bone mineral density. Quant Imaging Med Surg 2023;13:801-1. [Crossref] [PubMed]
12. Huang S, Cui X, Han H, Zhang Y, Gao B, Yu W. Study on the scanning protocols for measuring bone mineral density by gemstone CT spectral imaging based on European spine phantom. Acta Radiol 2023;64:346-52. [Crossref] [PubMed]
13. Ye H, Li X, Yao N, Shi Y, Wang Y, Yu W. Effect of abdominal adipose content on spine phantom bone mineral density measured by rapid kilovoltage-switching dual-energy CT and quantitative CT. Quant Imaging Med Surg 2022;12:4914-23. [Crossref] [PubMed]
14. Qin L, Huang J, Yu P, Yan J, Ge Y, Lu Y, Yan F, Wang L, Du L. Accuracy, agreement, and reliability of DECT-derived vBMD measurements: an initial ex vivo study. Eur Radiol 2021;31:191-9. [Crossref] [PubMed]
15. Li X, Li X, Li J, Jiao X, Jia X, Zhang X, Fan G, Yang J, Guo J. The accuracy of bone mineral density measurement using dual-energy spectral CT and quantitative CT: a comparative phantom study. Clin Radiol 2020;75:320.e9-320.e15. [Crossref] [PubMed]
16. Zhou S, Zhu L, You T, Li P, Shen H, He Y, Gao H, Yan L, He Z, Guo Y, Zhang Y, Zhang K. In vivo quantification of bone mineral density of lumbar vertebrae using fast kVp switching dual-energy CT: correlation with quantitative computed tomography. Quant Imaging Med Surg 2021;11:341-50. [Crossref] [PubMed]
17. Patino M, Prochowski A, Agrawal MD, Simeone FJ, Gupta R, Hahn PF, Sahani DV. Material Separation Using Dual-Energy CT: Current and Emerging Applications. Radiographics 2016;36:1087-105. [Crossref] [PubMed]
18. Xiao H, Hou P, Liu J, Gao J, Tan H, Liang P, Pu S. The best noise index combined with ASIR weighting selection in low-dose chest scanning. Zhonghua Yi Xue Za Zhi 2015;95:3041-4.
19. Li J, Gao J, Wang G, Cui D, Deng K. A phantom study using dual-energy spectral computed tomography imaging: Comparison of image quality between two adaptive statistical iterative reconstruction (ASiR, ASiR-V) algorithms for evaluating ground-glass nodules of the lung. J Cancer Res Ther 2021;17:1742-7. [Crossref] [PubMed]
20. Yang K, Cao J, Pisuchpen N, Kambadakone A, Gupta R, Marschall T, Li X, Liu B. CT image quality evaluation in the age of deep learning: trade-off between functionality and fidelity. Eur Radiol 2023;33:2439-49. [Crossref] [PubMed]
21. Mørup SD, Stowe J, Kusk MW, Precht H, Foley S. The impact of ASiR-V on abdominal CT radiation dose and image quality - A phantom study. J Med Imaging Radiat Sci 2022;53:453-9. [Crossref] [PubMed]
22. Jang H, Baek J. Method to estimate fan-beam CT noise power spectrum using two basis functions with a limited number of noise realizations. Med Phys 2022;49:1619-34. [Crossref] [PubMed]
23. Solomon J, Lyu P, Marin D, Samei E. Noise and spatial resolution properties of a commercially available deep learning-based CT reconstruction algorithm. Med Phys 2020;47:3961-71. [Crossref] [PubMed]
24. Zhao Y, Wu Y, Zuo Z, Cheng S. CT angiography of the kidney using routine CT and the latest Gemstone Spectral Imaging combination of different noise indexes: image quality and radiation dose. Radiol Med 2017;122:327-36. [Crossref] [PubMed]
25. Wang Y, Han H, Shi Y, Ye H, Yao N, Yu W. Effect of abdominal fat content on bone mineral density of lumbar body model measured by low milliampere quantitative CT. Chinese Journal of Osteoporosis 2023;29:173-8.