Assessing the repeatability and reproducibility of contrast time courses from a dynamic MRI flow phantom: initial results and experiences
Jacob M. Johnson1, Leah C. Henze Bancroft2, James H. Holmes3, Edward F. Jackson1,2, Frank R. Korosec1,2, Courtney K. Morrison2, Roberta M. Strigel1, Kang Wang3, and Ryan J. Bosca1

1Radiology, University of Wisconsin, Madison, WI, United States, 2Medical Physics, University of Wisconsin, Madison, WI, United States, 3GE Healthcare, Madison, WI, United States

### Synopsis

The recent development of a multi-modality, commercially available, dynamic flow phantom has provided a means of assessing the repeatability, reproducibility, and fidelity of contrast concentration time courses. In this work, we aimed to develop and evaluate a methodology for assessing the repeatability and reproducibility contrast concentration time courses derived from dynamic contrast-enhanced MR images of this dynamic flow phantom.

### Purpose

Dynamic contrast enhanced (DCE) MRI is increasingly being utilized as a clinical and research tool to derive quantitative imaging biomarkers of perfusion. While numerous efforts (e.g., those of the Radiological Society of North America’s Quantitative Imaging Biomarker Alliance and ISMRM Ad Hoc Committee on Standards for Quantitative MR) have been made to evaluate and mitigate the various sources of bias and variance in quantitative DCE-MRI studies, these groups have largely relied on static, multi-compartment phantoms that mimic the expected in vivo relaxivities [1,2]. The recent development of a multi-modality, commercially available, dynamic flow phantom [3,4] has provided a means of assessing the repeatability, reproducibility, and fidelity of contrast agent concentration time courses (CTC). While work to establish the repeatability and reproducibility of CTCs has been performed on CT scanners [4,5], similar work, to the best of the authors’ knowledge, has yet to be performed on MR. Therefore, the goal of this work was to develop and evaluate a methodology for assessing the repeatability and reproducibility of DCE-MRI derived CTCs using this dynamic flow phantom.

### Methods

The DCE perfusion phantom (Shelley Medical Imaging Technologies, London, ON) is represented schematically in Figure 1. To determine an optimal MR imaging configuration for quantifying CTCs, three imaging configurations were tested: (1) imaging the intended imaging block, (2) imaging the perfusion cylinder in the intended horizontal position, and (3) imaging the perfusion cylinder in an upright (vertical) position. An additional length of silicone tube was added at the exit of the pump to reduce the output pulsatility. The perfusion pump was set to a flow rate of 4mL/s and the control values were set to achieve equivalent flow between the cylinder and distribution output (see Figure 1). Prior to imaging, the pump was activated and given several minutes to stabilize. Images were acquired on two 1.5T clinical MRI system (Signa HDxt and Optima 450w, GE Healthcare, Waukesha, WI). All dynamic images were acquired with a 3D fast spoiled gradient echo sequence and temporal resolution less than 8 seconds. At approximately 90s into the acquisition, the contrast agent was injected using a power injector (5mL gadobenate dimeglumine followed by a 10mL saline flush delivered at 2mL/s). The optimal phantom configuration (as described in the results) was used to test the repeatability of the CTCs by acquiring three consecutive DCE scans during the same session, while the reproducibility was tracked by measurements made during three different scanning sessions. The concentration of contrast agent was estimated from the cylinder output (see Figure 2), and a gamma variate, given by $$C(t)=\kappa (t-t_0)^\alpha e^{\frac{t-t_0}{\beta}}$$ was fit to the resultant dynamic curve using QUATTRO [5]. The areas under the fitted curves (AUC) were calculated. Coefficients of variation (CV) were calculated for all gamma-variate shape parameters ($\alpha$ and $\beta$) and AUCs within a given scanning session (repeatability) and between all scanning sessions (reproducibility).

### Results

Practical considerations such as avoidance of flow artifacts, spatial/temporal resolution requirements, and pre-scanning difficulties resulted in more reliable CTC measurements derived from the output of the perfusion cylinder. Imaging of the cylinder in the upright (vertical) orientation (i.e., configuration 3) improved mixing and reduced the retention of contrast agent within cylinder, resulting in CTCs that were more consistent with the expected gamma-variate model. Therefore, repeatability and reproducibility data was acquired using configuration 3. All data exhibited good agreement with the fitted gamma-variate model (R2>0.988), an example of which is shown in Figure 3. The (repeatability) CVs calculated within a given scanning session for each of the three consecutive scans were: 4.2%, 12.3%, and 4.5% ($\alpha$); 5.8%, 5.6%, and 1.1% ($\beta$); and 11.9%, 2.5%, and 1.7% (AUC). The (reproducibility) CVs calculated from all acquisitions were: 18.4% ($\alpha$), 12.4% ($\beta$), and 25.4% (AUC).

### Conclusions

The CTC measurements with the DCE perfusion phantom arranged in configuration 3 were the most repeatable with maximum CVs reaching only 12%, while there was larger variability in longitudinal measurements (reproducibility). A number of potentially uncontrolled sources of variance require additional study, such as, MR hardware differences, phantom setup differences (i.e., different tube routes, inclusion/exclusion of the imaging block, etc.), flow ratio tuning, and power injector differences. With the high degree of repeatability observed, the DCE perfusion phantom could potentially be used as a means of validating the data fidelity of accelerated acquisition strategies used in the context of quantitative perfusion studies.

### Acknowledgements

The authors would like to acknowledge the support from the NIH (T32CA009206), the University of Wisconsin Department of Radiology R & D fund, RSNA, and GE healthcare.

### References

[1] R. Bosca, E. Ashton, G. Zahlmann, and E. F. Jackson, “RSNA Quantitative Imaging Biomarker Alliance (QIBA) DCE-MRI Phantom: Goal Design, and Initial Results,” 2012.

[2] K. E. Keenan, M. A. Boss, E. F. Jackson, S.-J. Kown, J. Dominique, and S. Russeck, “NIST/ISMRM MRI System Phantom T1 Measurements on Multiple MRI Systems,” in ISMRM 21st Annual Meeting & Exhibition, 2013.

[3] M. Peladeau-Pigeon and C. Coolens, “Computational fluid dynamics modelling of perfusion measurements in dynamic contrast-enhanced computed tomography: development, validation and clinical applications.,” Phys. Med. Biol., vol. 58, no. 17, pp. 6111–6131, 2013.

[4] B. Driscoll, H. Keller, and C. Coolens, “Development of a dynamic flow imaging phantom for dynamic contrast-enhanced CT,” Med. Phys., vol. 38, no. 8, p. 4866, 2011.

[5] B. Driscoll, H. Keller, D. Jaffray, and C. Coolens, “Development of a dynamic quality assurance testing protocol for multisite clinical trial DCE-CT accreditation.,” Med. Phys., vol. 40, no. 8, p. 081906, 2013.

[6] R. Bosca, V. Johnson, and E. Jackson, “QUATTRO: An Open-Source Software Package for Quantitative Imaging Applications in Assessing Treatment Response,” Med. Phys., vol. 41, no. 6, pp. 380–381, Jun. 2014.

### Figures

Figure 1. Perfusion pump schematic. The dashed boxes represent acquisition imaging volumes used for different acquisition configurations, while the numbers adjacent to the dashed boxes correspond to the imaging configuration.

Figure 2. Cylinder output contrast concentration time course measurement location. These images illustrate the contrast enhancement of a coronal acquisition of configuration 1 (left) and an axial acquisition of configuration 3 (right) with the corresponding ROIs used to derive the contrast agent time courses (red rectangles).

Figure 3. An example gamma-variate fit of the cylinder output. The red circles represent estimated gadolinium concentration ([Gd]) measured from the cylinder output. The blue and green lines, which are difficult to distinguish because of the close confidence intervals, represent the fitted model and confidence intervals, respectively.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
1476