Multiple centers are starting to use hyperpolarized 129Xe MR imaging and spectroscopy to quantify pulmonary function. This drives the need for standardization and quality assurance (QA), which is challenging for short-lived hyperpolarized agents. To address this, we developed a high pressure thermal 129Xe phantom constructed from high-density polyethylene and an associated loader shell that mimics the human torso. The phantom was imaged at 8 institutions across North America comprising a variety of scanners, vendors, coils, and field strengths. Image SNR measured with a minute-long scan agreed within a factor of 2 across sites and enabled rapid comparison of MR configurations.
Cylindrical thermal phantoms were constructed out of high-density polyethylene (HDPE) and rated to safely hold pressures of up to 300 psi. The phantoms have an internal volume of 5.85 L and were filled to 168.9 psia with a mixture of 61% natural xenon and 39% oxygen. The oxygen was used to shorten the T1 to 580 ms at 3T and 537 ms at 1.5T. A custom torso loader shell was constructed and filled with a saline mixture that mimicked typical RF-loading from human subjects. The phantom and loader shell are shown in Figure 1.
The scanner vendors, models, field strengths, and coils of the 8 centers are summarized in Figure 2 and the physical set-ups are shown in Figure 3. The recommended QA imaging sequence prescribed a non-slice selective coronal 2D gradient-echo (GRE) image acquired at the Ernst angle (74°) with a matrix size =64x32, FOV=440mmx440mm, TR/TE=750/6.13ms, BW=30Hz/pixel, and 4 averages (1.2 minute acquisition). An additional image was acquired without transmit voltage and TR=min to characterize noise. At each site, scan parameters were set to best mimic the recommended QA procedure with additional averages added, if necessary, to compensate for bandwidth limitations or field strength.
The phantom and noise images were reconstructed with minimal processing to best represent the true SNR of the MR system. The signal was defined by the mean pixel intensity in a hand-drawn ROI over the homogeneous portion of the phantom in the right side of the loader shell and the noise was quantified using the standard deviation of a large region of the noise scan as shown in Figure 4.
2D GRE Images were successfully obtained on all scanners, at all sites, with all available coils. A subset of these images with associated SNR values is shown in Figure 4. Across all these phantom scans, the SNR was 22.3 ± 5.4.
The best demonstration of a site-to-site comparison is between Duke and UVA (3.0T), for which the same pulse sequence and coil model were used. The SNR achieved at the two sites is almost identical (21.9 vs 20.5), suggesting strong reproducibility of phantom images between similar scanner platforms.
The phantom was also used to evaluate multiple coil configurations at McMaster and Robarts. The phantom was imaged with both a flexible vest coil and a rigid birdcage coil. This revealed that the birdcage coil appears to be slightly more sensitive than the flexible chest coil. The birdcage coil also provides a more homogeneous image that encompasses more of the phantom as shown in Figure 4.
As 129Xe MR imaging and spectroscopy gains wider utilization it becomes important to establish a mutually accepted standard for site-to-site comparison and regular QA. The high pressure 129Xe thermal phantom described here enabled an initial multi-site evaluation of system performance and demonstrated the potential for a universal QA imaging procedure.
This study revealed several challenges that prevented exactly duplicating the acquisition across three vendor platforms. For example, all GE scanners were limited to a bandwidth of approximately 90 Hz/pixel instead of the proposed 30 Hz/pixel, which was compensated by increasing the NEX by 3-fold to 12 compared to NEX=4 in the recommended protocol.
Varying difficulty was also experienced across all platforms in determining the reference voltage or transmit gain required to achieve the prescribed flip angle. In some cases voltage limits needed to be circumvented by increasing the pulse width. Moreover, it was found that many vendors include extensive filtering, and therefore, honest comparisons required off-line independent reconstruction of the images.
These challenges will be addressed by working with individual vendors to implement the recommended QA acquisition protocol. This should facilitate the congruence of acquisition and reconstruction techniques which are vital for ensuring accurate and comparable SNR values across platforms.