Introduction

T₁-weighted dynamic contrast enhanced (DCE)-MRI can be an important imaging tool for prediction and assessment of cancer treatment response. The repeatability of DCE-MRI has not been fully studied, particularly with contrast kinetic parameters including intracellular water lifetime (τ_i). We recently developed the 3D-UTE-GRASP (Golden angle Radial Sparse Parallel) method¹ for DCE-MRI study. With this sequence, datasets were acquired serially with isotropic resolution under high temporal resolution. The same protocol incorporates 3D-T₁ maps with identical resolution², enabling quantitative analysis of a whole tumor³. The objective of this study was to investigate the repeatability of the proposed DCE-MRI technique for contrast kinetic parameters including τ_i.

Methods

Six to eight-week-old C57BL6 mice (n=7) with GL261 mouse glioma models were included in this study. MRI experiments were performed on a Bruker 7T micro-MRI system, with a ¹H four-channel phased array cryogenically cooled receive-only MRI coil. Two DCE-MRI experiments were carried out with a 30 minutes gap. DCE-MRI image acquisition was performed using 3D-UTE-GRASP pulse sequence¹ (TR=4ms and TE=0.028ms) to achieve an isotropic spatial resolution and to minimize the T₂* effect. It was continuously run to acquire 154,080 spokes (51,360 spokes each flip angle segment 8^o - 25^o - 8^o) for 10 minutes and 13 seconds. Reconstruction temporal frame resolution was T = 5 s/frame. Image matrix = 128x128x128, field of view = 20x20x20 mm³ and the spatial resolution was 0.156x0.156x0.156 mm³. A bolus of gadoxetate disodium (Eovist, Bayer) in saline at the dose of 0.1 mmol/kg was injected through a tail vein catheter, starting 60 seconds after the start of data acquisition. Prior to each DCE-MRI experiment, a 3D T₁ map with the same isotropic high resolution was obtained using the same 3D-UTE-GRASP sequence² with variable flip angles (8^o - 2^o - 12^o, 12,776 spokes for each flip angle, with total acquisition time was 153 s). To align the tumor voxels between the 1^st and the 2^nd DCE-MRI scans, the 2^nd experiment (both T₁ and DCE-MRI) was registered to the 1^st DCE-MRI by applying rigid body transformation (Figure 1A and 1B). Arterial input function (AIF) was obtained using the Principal Component Analysis (PCA) method used in our previous study⁴ with the measured T₁ maps described above (Figure 1C). Pharmacokinetic model analysis was carried out for the tumor with the two-compartment exchange model with water exchange⁴, yielding five parameters: PS (permeability surface area product), F_p (blood flow), v_e (extracellular space volume fraction), v_p (vascular space volume fraction) and τ_i (intracellular water lifetime) (Figure 2).

Results and Discussions

Figure 2 shows that v_e and v_p showed small differences between the 1^st and 2^nd scans. In contrast, PS and F_p appear to be higher in the 2^nd scan. τ_i showed a large variability between the 1^st and 2^nd experiments. Figure 3 shows the boxplots comparison of the whole tumor pharmacokinetic model parameters. Figure 4 shows comparisons of the median values of all 7 animals. Similar to the observation in Figure 2, v_e and v_p showed strong correlations between the two scans, with correlation coefficients were 0.61 and 0.70, respectively. Wilcoxon signed rank test showed that there is not enough evidence to reject the null hypothesis of equal median with p-values of 0.375 (v_e) and 0.467 (v_p). This is reasonable since cellular volumetric properties were expected to remain same within one hour of the two scans. The other two parameters, PS and F_p, of the 2^nd experiment seem to be higher than those of the 1^st scan, which may indicate the actual changes of the physiological properties of the tumor between two scans. The last parameter, τ_i, showed a large variability between the two scans. Wilcoxon signed rank test showed the failure to reject the null hypothesis with p-values are 0.109 (PS), 0.078 (F_p) and 0.156 (τ_i) respectively. Only the p-value for K^trans (0.047) showed that the rejection of the null hypothesis. Figure 5 shows the comparison of whole tumor τ_i estimation precision (evaluated with interquartile range, IQR) changes with K^trans changes. Out of six animals with K^trans increases in the 2^nd scan, four animals showed improved precision in τ_i estimation (IQR decreases in the 2^nd scan). One of the limitations of the current study is the repetition of the two DCE-MRI experiments within one hour, where the contrast agent from the 1^st DCE-MRI was still not fully washed out (which can be observed from the relative lower T₁ values in Figure 1A) during the 30 min gap between the two scans.

Conclusion

This study demonstrates the feasibility and limitation of using the 3D-UTE-GRASP sequence based DCE-MRI technique for longitudinal studies with pharmacokinetic model analysis. While the DCE-MRI parametric measurements were found consistent through the entire tumor volumes between two scans, our results showed that F_p and PS could change between two consecutive scans that could also affect the estimation of τ_i.

Acknowledgements

NIH R01CA160620, NIH R01CA219964, P41EB017183, NIH/NCI 5P30CA016087