Pancreatic cancer is highly desmoplastic and slowly takes up extracellular gadolinium based contrast agent during MR imaging. Quantitative estimation of gadolinium based contrast uptake in pancreas cancers may help assess the tumor stroma, which is implicated in tumor aggressiveness and treatment response. However, tissue motion and sensitivity inhomogeneity on abdominal MR scans present complications for quantitative analysis of contrast uptake. The method demonstrated here relies on fast imaging and data-driven analysis to make the quantification of contrast agent concentration in pancreatic cancer more robust against patient motion and sensitivity variation.All images were acquired on a GE MR-PET 3T system at UCSF using an embedded body transmitter with a Central Matrix Array receiver (GE Healthcare, Waukesha WI). Quantification of contrast agent concentration requires measurement of the change in T1 relaxation time of the tissue due to the arrival of the agent. Images were acquired using a fast multiecho 3D gradient echo imaging sequence. Measurement of the T1 relaxation time was performed by measuring 5 different flip angles using the Variable Flip Angle (VFA) [1] method after registering each acquisition to one chosen as a reference. Volumes were also acquired prior to and at 3 minutes, 5 minutes, 7.5 minutes, and 10 minutes following bolus injection of a gadolinium contrast agent (Gadavist, Bayer). Images were transferred offline in DICOM format, anonymized, and analyzed. Total imaging time was roughly 1 hour, including all MR acquisitions and simultaneous PET acquisitions using 18FDG. Each volume was acquired during a single breath hold, requiring roughly 30 seconds to cover a 400 mm by 400 mm by 435 mm volume with 256x256x128 resolution. While the breath hold and fast acquisition are sufficient to minimize the effects of motion during the scan, motion between scans can only be corrected by registration. To differentiate between fat and water a multi echo acquisition (FLEX) was used, and the water image was calculated as a weighted average of two of the echoes. The pre-contrast T₁ and the signal change due to the contrast agent allow estimation of the contrast agent concentration. If the contrast agent concentration is sufficiently low and the echo time is short, the T₂^* effect is a small correction, neglected by many authors [1]. E₁ is obtained from the pre contrast VFA acquisitions at 3, 6, 8, 10, and 15 degrees. By assumption the flip angles were all uniform and correct across the abdomen. The system operated in circularly polarized mode with a preset phase offset between the two drive channels which was not optimized for each patient; by inspection, none of the images showed significant quadrupole artifacts. The 6 degree VFA dataset was chosen as a standard, and all other volumes were registered to it using a rigid 6 degree of freedom alignment followed by an affine transformation using Slicer. A binary threshold mask was also created in Slicer to ensure that the fit was only performed in voxels containing significant signal.The precontrast signal is given by $$ S=\frac{M_0E_2^*\sin\alpha}{1-E_1\cos\alpha} $$ and similarly the post-contrast signal by $$ S'=\frac{M_0E_2^{'*}\sin\alpha}{1-E_1'\cos\alpha}$$ Defining $$ \beta=\frac{E_2^*}{E_2^{'*}}$$ one finds that $$ E_1' = \frac{1-\frac{S}{S'}\frac{1-E_1\cos\alpha}{\beta}}{\cos\alpha} $$ β and E₁' are both functions of contrast agent concentration making the fit more complicated, requiring either an iterative approach or some simplifying assumption. A further complication is the use of the multi echo acquisition, since the water image used here is a weighted average of two different echoes acquired at different echo times. Here we set β=1, resulting in an overestimate of the absolute contrast agent concentration. Given E₁^' and E₁, [Gd] was calculated using the sequence repetition time and the effective T₁ relaxivity of Gadavist at 3T, taken from the literature [3].Example slices from the pre-contrast images, T₁ maps, and relative concentration maps are shown below for one patient. Good data quality is obtained throughout the liver and pancreas, and significant heterogeneity within the tumor is apparent.The methods demonstrated here allow the collection of high quality maps of relative concentration agent uptake in the liver and pancreas in pancreatic cancer patients during a simultaneous MR-PET study, proving the feasibility of the techniques. Better quantitation of the T₂^* changes accounting for the effect of the FLEX acquisition will allow better estimation of β and so better estimation of the contrast agent concentration and therefore better estimation of the extracellular [Gd] and extracellular volume fraction, a potential biomarker for tumor stroma.