# Potential for Dose Reduction in CT-Derived Left Ventricular Ejection Fraction: A Simulation Study

^{1}

^{2}

^{3}

^{4}

^{5}

^{*}

## Abstract

**:**

## 1. Introduction

## 2. Materials and Methods

#### 2.1. Simulation Data

- Helical CCTAs performed without ECG-based dose reduction, including ED and ES phases.
- Maximum phase interval equal to or less than 10% of the cardiac cycle.

- LV contrast attenuation below 250 HU, defined as minimally acceptable by SCCT guidelines [13]. ROI measurement was performed on a single axial slice in the ED phase, midway between the mitral valve and apex.
- Severe cardiac motion or respiratory artefacts with distortion of endocardial contours.
- Metal implant artefacts, e.g., from pacemaker leads, metallic aortic valves, or thoracic spine implants.
- Excessive noise. Studies were excluded if the PACS report noted compromised image quality related to noise. Otherwise, the study was evaluated subjectively by the first author, who has over 17 years of experience in cardiac CT.
- Failure to reliably identify ED and ES phases in the time–volume curves as clearly distinguishable maximum and minimum volumes.

#### 2.2. Optimisation of Simulation Parameters

**Figure 1.**Visualisation of the testing loop used to identify parameters for the maximum similarity between the real and simulated images. Circles represent parameters adjusted at each step. The maximum dose (300 mAs) image was used as input for 13 simulated dose levels for each parameter combination. The contrast-to-noise ratio (CNR) between the myocardium and LV cavity and the SSIM between real and simulated images were calculated and compared for each dose level.

**Figure 2.**Placement of the two ROIs for contrast-to-noise ratio (CNR) calculation. The phantom is masked with the green ROI, ensuring that the structural similarity index (SSIM) is only calculated for the heart. Blue and red circles are the ROIs used to measure attenuation and noise in the LV and myocardium, respectively.

^{2}value were calculated in each case. A slope of one is not necessarily predictive of CNR similarity if individual measurements are scattered about the prediction line, resulting in low R

^{2}. Likewise, a perfect R

^{2}of one but a slope significantly different from one would mean systematic over- or underestimation of simulated CNR.

_{SSIM}) was calculated for each parameter combination (Figure 3). We masked the phantom contours with a polygonal ROI and thereby excluded other phantom tissue from SSIM calculation (Figure 2).

^{2}) from CNR linear regression of simulated vs. real dose-reduced images and AUC

_{SSIM}were normalised to a scale from zero to one. A measure of deviation between real and simulated images was defined from these three metrics. This was defined as the vector length (norm) of the three components. It is shown in Equation (2), where the subscript “norm” indicates normalised values.

**Figure 3.**SSIM between simulated and real exposure images at all 13 dose levels. In this particular case, stratified by reconstruction filter, but for fixed values of sampling interval (0.6 degrees), default photon flux (10

^{5}) and filter frequency cutoff (80%).The area under the SSIM vs dose curve (AUC

_{SSIM}) is then calculated by integrating SSIM over the entire dose range for each parameter combination.

#### 2.3. Image Processing

^{3}and 10

^{7}. A circular ROI was drawn in the LV cavity, avoiding PM, while a polygonal ROI was drawn on the lateral myocardium, avoiding visible coronaries. The image was then processed at all 200 initial flux values. ROI SD was measured automatically at all levels, and noise from the corresponding image without noise insertion was subtracted. The data were smoothed using a Gaussian kernel and the minimal difference for each ROI was automatically identified. The mean of the corresponding flux values was then passed to the noise insertion algorithm, and the study reconstructed at five dose levels: 100, 25, 10, 5 and 2% relative dose.

#### 2.4. Image Analysis

_{vol}) and dose-length product were recorded. Effective dose was calculated using a conversion factor of 0.026 mSv/mGycm [19].

#### 2.5. Statistics

## 3. Results

## 4. Discussion

_{vol}was 2.3 mGy, and conservatively rounding up to 2.5 mGy would be our suggestion when programming the ATCM. A clinical protocol should be validated against an external reference such as short-axis cardiac cine MRI, which is complicated by intrinsic modality differences (primarily spatial and temporal resolution) [3,24], hemodynamic effects of contrast media [25] and different measurement software [26]. Ideally, before a clinical study, our simulation could be applied to a dataset of full-dose functional CT images with a corresponding reference modality. Such an approach could help disentangle the effects of the differences mentioned above from that of noise introduced by dose reduction.

#### Limitations

_{vol}values indicates a similar diversity in patient sizes. There was a substantial variation of CTDI

_{vol}values in the included scans, likely due to different patient characteristics and indications. The range is comparable to that reported by [21] when comparing doses from different hospitals, and it is believed that our cohort can be considered a representative sample of a northern European population.

## 5. Conclusions

_{vol}of 2.5 mGy without significantly affecting LVEF. If results can be validated in a prospective study, CT may be a viable alternative to MUGA for serial monitoring of LVEF, with a large cumulative dose reduction.

## Author Contributions

## Funding

## Institutional Review Board Statement

## Informed Consent Statement

## Data Availability Statement

## Conflicts of Interest

## Abbreviations

ALARA | As Low as Reasonably Achieveable |

ATCM | Automatic Tube Current Modulation |

ATVS | Automatic Tube Voltage selection |

AUC | Area Under the Curve |

CTDI | Computed Tomography Dose Index |

CI | Confidence Interval |

DLP | Dose–Length Product |

ED(V) | End-Diastolic (Volume) |

ES(V) | End-Systolic (Volume) |

LVEF | Left Ventricular Ejection Fraction |

SSIM | Structural Similarity Index |

## References

- Stone, J.R.; Kanneganti, R.; Abbasi, M.; Akhtari, M. Monitoring for Chemotherapy-Related Cardiotoxicity in the Form of Left Ventricular Systolic Dysfunction: A Review of Current Recommendations. JCO Oncol. Pract.
**2021**, 17, 228–236. [Google Scholar] [CrossRef] [PubMed] - Van Dijk, J.D. Dose-optimization in nuclear cardiac imaging, time for the next step? J. Nucl. Cardiol.
**2019**, 26, 1981–1983. [Google Scholar] [CrossRef] [PubMed] - Pickett, C.A.; Cheezum, M.K.; Kassop, D.; Villines, T.C.; Hulten, E.A. Accuracy of cardiac CT, radionucleotide and invasive ventriculography, two- and three-dimensional echocardiography, and SPECT for left and right ventricular ejection fraction compared with cardiac MRI: A meta-analysis. Eur. Heart J. Cardiovasc. Imaging
**2015**, 16, 848–852. [Google Scholar] [CrossRef] - Kaniewska, M.; Schuetz, G.; Willun, S.; Schlattmann, P.; Dewey, M.; Schuetz, G.M. Noninvasive evaluation of global and regional left ventricular function using computed tomography and magnetic resonance imaging: A meta-analysis. Eur. Radiol.
**2017**, 27, 1640–1659. [Google Scholar] [CrossRef] [PubMed] - Hell, M.; Steinmann, B.; Scherkamp, T.; Arnold, M.; Achenbach, S.; Marwan, M. Analysis of left ventricular function, left ventricular outflow tract and aortic valve area using computed tomography: Influence of reconstruction parameters on measurement accuracy. Br. J. Radiol.
**2021**, 94, 20201306. [Google Scholar] [CrossRef] [PubMed] - Choi, Y.; Ahlman, M.; Mallek, M.; Cork, T.; Chen, M.; Bluemke, D.; Sandfort, V. Cardiac cine CT approaching 1 mSv: Implementation and assessment of a 58-ms temporal resolution protocol. Int. J. Cardiovasc. Imaging
**2020**, 36, 1583–1591. [Google Scholar] [CrossRef] - Lee, J.W.; Nam, K.J.; Kim, J.Y.; Jeong, Y.J.; Lee, G.; Park, S.M.; Lim, S.J.; Choo, K.S. Simultaneous Assessment of Left Ventricular Function and Coronary Artery Anatomy by Third-generation Dual-source Computed Tomography Using a Low Radiation Dose. J. Cardiovasc. Imaging
**2020**, 28, 21. [Google Scholar] [CrossRef] - Groves, D.W.; Olivieri, L.J.; Shanbhag, S.M.; Bronson, K.C.; Yu, J.H.; Nelson, E.A.; Rollison, S.F.; Stagliano, M.S.; John, A.S.; Kuehl, K.; et al. Feasibility of low radiation dose retrospectively-gated cardiac CT for functional analysis in adult congenital heart disease. Int. J. Cardiol.
**2017**, 228, 180–183. [Google Scholar] [CrossRef] [PubMed] - Lesser, A.M.; Newell, M.C.; Samara, M.A.; Gornick, C.; Grant, K.; Garberich, R.; Han, B.K. Radiation dose and image quality of 70 kVp functional cardiovascular computed tomography imaging in congenital heart disease. J. Cardiovasc. Comput. Tomogr.
**2016**, 10, 173–178. [Google Scholar] [CrossRef] - Pellikka, P.A.; She, L.; Holly, T.A.; Lin, G.; Varadarajan, P.; Pai, R.G.; Bonow, R.O.; Pohost, G.M.; Panza, J.A.; Berman, D.S.; et al. Variability in Ejection Fraction Measured By Echocardiography, Gated Single-Photon Emission Computed Tomography, and Cardiac Magnetic Resonance in Patients with Coronary Artery Disease and Left Ventricular Dysfunction. JAMA Netw. Open
**2018**, 1, e181456. [Google Scholar] [CrossRef] - Plumhans, C.; Keil, S.; Ocklenburg, C.; Mühlenbruch, G.; Behrendt, F.; Günther, R.; Mahnken, A. Comparison of manual, semi- and fully automated heart segmentation for assessing global left ventricular function in multidetector computed tomography. Investig. Radiol.
**2009**, 44, 476–482. [Google Scholar] [CrossRef] [PubMed] - Wai, B.; Thai, W.; Brown, H.; Truong, Q. Novel phase-based noise reduction strategy for quantification of left ventricular function and mass assessment by cardiac CT: Comparison with cardiac magnetic resonance. Eur. J. Radiol.
**2013**, 82, e337–e341. [Google Scholar] [CrossRef] [PubMed] - Abbara, S.; Blanke, P.; Maroules, C.D.; Cheezum, M.; Choi, A.D.; Han, B.K.; Marwan, M.; Naoum, C.; Norgaard, B.L.; Rubinshtein, R.; et al. SCCT guidelines for the performance and acquisition of coronary computed tomographic angiography: A report of the society of Cardiovascular Computed Tomography Guidelines Committee: Endorsed by the North American Society for Cardiovascular Imaging (NASCI). J. Cardiovasc. Comput. Tomogr.
**2016**, 10, 435–449. [Google Scholar] [CrossRef] - Biguri, A.; Dosanjh, M.; Hancock, S.; Soleimani, M. TIGRE: A MATLAB-GPU toolbox for CBCT image reconstruction. Biomed. Phys. Eng. Express
**2016**, 2, 055010. [Google Scholar] [CrossRef] - Pelt, D.M.; Batenburg, K.J. Accurately approximating algebraic tomographic reconstruction by filtered backprojection. In Proceedings of the 13th International Meeting on Fully Three-Dimensional Image Reconstruction in Radiology and Nuclear Medicine, Newport, RI, USA, 1–4 June 2015; pp. 158–161. [Google Scholar]
- Kusk, M.; Stowe, J.; Hess, S.; Gerke, O.; Foley, S. Low-cost 3D-printed anthropomorphic cardiac phantom, for computed tomography automatic left ventricle segmentation and volumetry—A pilot study. Radiography
**2023**, 29, 131–138. [Google Scholar] [CrossRef] [PubMed] - Samei, E.; Krupinski, E. The Handbook of Medical Image Perception and Techniques, 2nd ed.; Cambridge University Press: Cambridge, UK, 2018. [Google Scholar]
- Wang, Z.; Bovik, A.C.; Sheikh, H.R.; Simoncelli, E.P. Image quality assessment: From error visibility to structural similarity. IEEE Trans. Image Process
**2004**, 13, 600–612. [Google Scholar] [CrossRef] - Trattner, S.; Halliburton, S.; Thompson, C.; Xu, Y.; Chelliah, A.; Jambawalikar, S.; Peng, B.; Peters, M.; Jacobs, J.; Ghesani, M.; et al. Cardiac-Specific Conversion Factors to Estimate Radiation Effective Dose From Dose-Length Product in Computed Tomography. JACC Cardiovasc. Imaging
**2018**, 11, 64–74. [Google Scholar] [CrossRef] - Mao, S.S.; Li, D.; Rosenthal, D.G.; Cerilles, M.; Zeb, I.; Wu, H.; Flores, F.; Gao, Y.; Budoff, M.J. Dual-standard reference values of left ventricular volumetric parameters by multidetector CT angiography. J. Cardiovasc. Comput. Tomogr.
**2013**, 7, 234–240. [Google Scholar] [CrossRef] - Stocker, T.J.; Leipsic, J.; Hadamitzky, M.; Chen, M.Y.; Rubinshtein, R.; Deseive, S.; Heckner, M.; Bax, J.J.; Kitagawa, K.; Marques, H.; et al. Application of Low Tube Potentials in CCTA: Results From the PROTECTION VI Study. JACC Cardiovasc. Imaging
**2020**, 13, 425–434. [Google Scholar] [CrossRef] - Söderberg, M. Overview, Practical Tips and Potential Pitfalls of Using Automatic Exposure Control in CT: Siemens Care Dose 4D. Radiat. Prot. Dosim.
**2016**, 169, 84–91. [Google Scholar] [CrossRef] [PubMed] - Li, W.; Diao, K.; Wen, Y.; Shuai, T.; You, Y.; Zhao, J.; Liao, K.; Lu, C.; Yu, J.; He, Y.; et al. High-strength deep learning image reconstruction in coronary CT angiography at 70-kVp tube voltage significantly improves image quality and reduces both radiation and contrast doses. Eur. Radiol.
**2022**, 32, 2912–2920. [Google Scholar] [CrossRef] - Clark, J.; Ionescu, A.; Chahal, C.A.A.; Bhattacharyya, S.; Lloyd, G.; Galanti, K.; Gallina, S.; Chong, J.H.; Petersen, S.E.; Ricci, F.; et al. Interchangeability in Left Ventricular Ejection Fraction Measured by Echocardiography and cardiovascular Magnetic Resonance: Not a Perfect Match in the Real World. Curr. Probl. Cardiol.
**2023**, 48, 101721. [Google Scholar] [CrossRef] [PubMed] - John, A.M.; Yadav, S. Effect of bolus administration of non-ionic radiopaque contrast media on blood pressure variation. Radiography
**2019**, 25, 346–348. [Google Scholar] [CrossRef] - Moody, W.E.; Hudsmith, L.E.; Holloway, B.; Treibel, T.A.; Davies, R.; Kozor, R.; Hamilton-Craig, C.; Edwards, N.C.; Bradlow, W.M.; Moon, J.C.; et al. Variation in cardiovascular magnetic resonance myocardial contouring: Insights from an international survey. J. Magn. Reson. Imaging
**2019**, 50, 1336–1338. [Google Scholar] [CrossRef] [PubMed] - Muenzel, D.; Koehler, T.; Brown, K.; Zabic, S.; Fingerle, A.; Waldt, S.; Bendik, E.; Zahel, T.; Schneider, A.; Dobritz, M.; et al. Validation of a low dose simulation technique for computed tomography images. PLoS ONE
**2014**, 9, e107843. [Google Scholar] [CrossRef] [PubMed]

**Figure 4.**Scatterplot of phantom myocardium to left ventricle CNR, with linear regression fit for all 13 real/simulated dose levels. Note scale differences due to different noise characteristics of the two kernels (Br32 and Bv40). The dashed line represents the line of perfect correspondence.

**Figure 5.**Clinical image reconstructed at five simulated dose levels (in end-diastolic phase). The simulated dose level is indicated in yellow font.

**Figure 6.**Left ventricle segmentation at 100, 10 and 2% of the initial dose. The red contours are the automatic LV endocardium segmentation. Red-coloured pixels are those above the threshold for LV volume calculation in blood volume mode. The green countours are LV epicardium segmentation, not used in LVEF calculation.

**Figure 7.**Bland-Altman plots of LVEF for each simulated dose level vs. 100% dose images. The black dashed line is the mean bias, while the blue lines represent the upper and lower limits of agreement. Double arrows are 95% CI of estimates.

Parameter | Value (s) |
---|---|

Collimation [mm] | 2 × 96 × 0.6 |

Tube voltage [kVp] | 120 |

Tube time–current range [mAs] | 10–300 |

Pitch | 0.25 |

Slice thickness/increment [mm] | 1.0/0.8 |

Matrix | 256 × 256 |

Kernel/ADMIRE level | Br32/5, Bv40/3 |

Scan range [mm] | 160 |

**Table 2.**Different parameter combinations tested for simulation parameter optimisation. The angular sampling interval is for forward-projecting to sinogram space. Initial photon flux describes the photon flux corresponding to 100% dose. The reconstruction filter and the associated spatial cutoff are equivalent to the FBP reconstruction kernel.

Initial Photon Flux [×10^{3}] | Sampling Interval [degrees] | Filters | Cutoff [%] |
---|---|---|---|

1 | 0.4 | Ram-Lak | 20 |

5 | 0.5 | Shepp-Logan | 40 |

10 | 0.6 | Cosine | 60 |

50 | 0.7 | Hamming | 80 |

100 | 0.8 | Hann | 100 |

500 | 0.9 | ||

1000 | 1.0 | ||

5000 | |||

10,000 |

**Table 3.**Characteristics of studies included for noise simulation. Height/Weight was only available in three cases, so it was not tabulated.

Sex [M/F] | Median Age [years] (Min-Max) | Mean DLP [mGycm] (SD) | Mean CTDI [mGy] (SD) | Mean Eff. Dose [mSv] (SD) |
---|---|---|---|---|

27/24 | 73 (31–92) | 771.9 (545.8) | 47.1 (31.6) | 20.1 (14.7) |

**Table 4.**Optimal parameters for both kernels and their metrics as determined in the iterative testing loop.

Original kernel | Br32 | Bv40 |
---|---|---|

Filter | Hamming | Hamming |

Angular sampling [degrees] | 0.6 | 0.4 |

Filter cutoff [%] | 0.6 | 0.8 |

Default photon flux | 5 × 10^{5} | 10^{5} |

Normalised AUC_{SSIM} | 0.998 | 0.994 |

CNR regression slope | 1.011 | 1.003 |

CNR R^{2} | 0.962 | 0.978 |

**Table 5.**LVEF bias and upper/limits-of agreement in comparison with 100% dose images. Figures in parentheses are 95% CI of point estimates. p-values resulting from paired t-test. “BV” and “ST” represent blood volume and standard mode, respectively.

Dose Level | Bias | p-Value | LoA− | LoA+ |
---|---|---|---|---|

BV | ||||

25% | 0.0 (−0.6,0.7) | 0.905 | −4.5 (−5.9,−3.6) | 4.6 (3.7,5.9) |

10% | 0.6 (−0.7,1.9) | 0.368 | −8.5 (−11.1,−6.6) | 12.2 (9.9,15.3) |

5% | 1.4 (−0.2,2.9) | 0.081 | −9.4 (−12.5,−7.2) | 12.2 (9.9,15.3) |

2% | 4.4 (2.3,6.5) | <0.001 | −10.2 (−14.4,−7.1) | 19.0 (16.0,23.2) |

ST | ||||

25% | 1.5 (0.8,2.1) | <0.001 | −2.9 (−4.2,−2.0) | 5.9 (5.0,7.2) |

10% | 2.2 (1.3,3.0) | <0.001 | −3.8 (−5.5,−2.6) | 8.1 (6.9,9.9) |

5% | 3.0 (2.0,4.0) | <0.001 | −4.1 (−6.1,−2.6) | 10.0 (8.5,12.0) |

2% | 6.0 (4.2,7.8) | <0.001 | −6.5 (−10.1,−3.9) | 18.5 (15.9,22.1) |

**Table 6.**Mean ESV and EDV for the different simulated dose levels and both measurement methods. Numbers in parentheses are standard deviations. Asterisks indicate that the mean value is significantly different from the 100% dose level (p < 0.05).

Dose Level | EDV [mL] | ESV [mL] | ||
---|---|---|---|---|

BV | ST | BV | ST | |

100% | 126.5 (40.0) | 151.9 (45.2) | 44.3 (22.5) | 58.7 (28.7) |

25% | 126.3 (42.3) | 149.4 (45.5) | 44.8 (23.0) | 56.3 (28.7) * |

10% | 130.6 (42.3) * | 149.9 (47.6) | 44.8 (23.2) | 54.8 (28.3) * |

5% | 129.9 (42.3) | 147.7 (47.3) | 43.5 (23.2) | 52.6 (27.4) * |

2% | 127.6 (46.1) | 147.3 (53.0) | 37.8 (20.0) * | 46.4 (22.7) * |

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |

© 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Kusk, M.W.; Hess, S.; Gerke, O.; Foley, S.J.
Potential for Dose Reduction in CT-Derived Left Ventricular Ejection Fraction: A Simulation Study. *Tomography* **2023**, *9*, 2089-2102.
https://doi.org/10.3390/tomography9060164

**AMA Style**

Kusk MW, Hess S, Gerke O, Foley SJ.
Potential for Dose Reduction in CT-Derived Left Ventricular Ejection Fraction: A Simulation Study. *Tomography*. 2023; 9(6):2089-2102.
https://doi.org/10.3390/tomography9060164

**Chicago/Turabian Style**

Kusk, Martin Weber, Søren Hess, Oke Gerke, and Shane J. Foley.
2023. "Potential for Dose Reduction in CT-Derived Left Ventricular Ejection Fraction: A Simulation Study" *Tomography* 9, no. 6: 2089-2102.
https://doi.org/10.3390/tomography9060164