INTRODUCTION: Genetic engineered mouse (GEM) model of pancreatic ductal adenocarcinoma (PDA) provides a powerful tool for developing effective treatment strategy for this deadly cancer. As a modality sensitive to tumor microenvironment (TME), diffusion weighted (DW)-MRI might be useful for assessing tumor responses to stroma-directed drugs, which degrade or reduce extracellular matrix components in PDA. Fast respiration rate of mice combined with high magnetic field (>3T) and less advanced hardware of preclinical scanner impose a challenge for quantitative DW-MRI of abdominal cancers in mice including liver and pancreatic tumor. Radial k-space sampling intrinsically resists motion while echo-planar imaging (EPI) freezes motion via fast imaging and employed in clinical DW-MRI. However, their ability to minimize motion artefacts have not been evaluated in murine orthotopic PDA model. Employing a GEM model of PDA, we designed a test retest study to compare the two techniques side-by-side for accuracy and reproducibility as well as motion-robustness.
METHODS: Genetic engineered mouse (GEM) that carries Kras and p53 mutation in pancreas specific Cre allele referred to as KPC model ¹ was used. MRI was performed on 4.7T horizontal bore DirectDrive® MR system (Agilent) interfaced with a 12-cm gradient coil. Respiration and core temperature were monitored (SAI Inc, Stonybrook, NY) and the core temperature was maintained at 37±0.2°C by directing warm air into the bore of the magnet. The mouse was placed on top of two 5-mm NMR tubes containing water and butanol, respectively. The test-retest study enrolled KPC mice bearing PDA tumor in the range of 138-929 mm³. After the test study, the mouse was recovered from anesthesia and returned to the cage for 2 hours before re-anesthetized. The RF coil was retuned and recalibrated.
Both Radial and EPI DW-MRI protocol employ a multi-slice 2D spin echo (SE)-DW preparation ²: diffusion sensitizing gradients with a separation time (∆) of 0.0144 sec applied for a duration (δ) of 0.009 sec to +x, +y, +z and then -x, -y, -z with varied amplitude, resulting in five b-values (0.64, 535, 1071, 1478, 2141 s/mm²); FOV=32 x 32 mm². Radial-DW acquisition includes 256 views (spokes), each having 128 points; one signal average. For EPI-DW, the entire k-space (128 phase encoding lines) was sampled in 4 shots, each acquiring 32 echoes per TR of 250 msec; 16 signal averages were used to obtain similar SNR as Radial-DW images. Abdominal images of mice acquired by EPI-DW protocol with 1- or 2-shot have unacceptable SNR. Respiration gated and non-gated acquisition were compared in both protocols. Test-retest statistics including within-subject coefficient-of-variation (CV_WS) are computed ^3,4.
RESULTS: Due to motion susceptible location of pancreatic tumor, respiration-gating implemented with SE-DW protocol does not adequately suppress respiratory motion artefacts (Fig 1A) therefore additional motion resistant mechanism is necessary for murine abdominal DW-MRI. Both Radial-DW and EPI-DW protocol (with respiration gating) lead to motion-free diffusion-weighted images even at very high b-values (Fig 1B-1C). Comparing to Radial-DW, EPI-DW results in ghosting artefacts around the butanol phantom (white arrows in Fig 1C). The ADC of water phantom measured by radial-DW is almost identical to literature value of 3.2 x 10-3 mm²/s at 37°C ⁵.
Bland-Altman plots, where ADC values of water phantom measured by Radial or EPI-DW protocol were compared with the literature value are shown in Fig 2: in plot derived from Radial-DW (2A), bias is minimal and ±95% CI (confidence interval) lines are evenly distributed around x-axis; In contrast, in plot from EPI-DW (2B), equality line is outside the ±95% CI limit, suggesting a systemic error introduced by EPI-DW. Since all data points are distributed on the left of the literature value (dotted line), EPI-DW consistently underestimates water ADC.
Statistics for test-retest studies (Table-1) show that CV_WS of tumor and muscle ADC <10% achieved by Radial-DWI, suggesting an excellent test-retest reproducibility for a motion susceptible tumor compared to CV_WS of 15% by EPI-DW. ADC values of pancreatic tumor measured by Radial- and EPI-DW are highly correlated (r² =0.82, Fig 3A). Bland-Altman plot of test-retest ADC values revealed that Radial-DW leads to a smaller bias and better agreement (smaller width between ±95% CI) than ADC from EPI-DW protocol (Fig 3B-3C).
DISCUSSION: Both Radial- and EPI-DW protocol (in combination with respiration gating) can effectively remove motion artefacts from DW images although ghosting persists in EPI-DW images likely due to multi-shot k-space sampling. A systemic error in ADC of water (reference to literature) is only observed with EPI-DW but not with Radial-DW protocol. Furthermore, Radial-DW leads to better test-retest producibility with lower CV_WS compared to EPI-DW.
CONCLUSION: Our data revealed that motion-robustness for DW-MRI of orthotopic pancreatic tumor in mice can be obtained by Radial-DW protocol with high degree of accuracy and good test-retest reproducibility. Out study is consistent with recent QIBA recommendations for improved precision of DW-MRI biomarkers for oncology trial ³.

Acknowledgements

This study was partially supported by U24CA231858 (Penn Quantitative Imaging Resource for Pancreatic Cancer), R21CA198563 and R01CA211337.