Introduction

Quantitative precise measurements of the dynamic arterial blood concentration (Arterial Input Function, AIF) are challenging in perfusion MRI, especially in small animals. However, quantitative precision of the AIF directly impacts parameters derived with pharmacokinetic modeling (PKM). In clinical and preclinical Positron Emission Tomography (PET), extracorporeal recordings of the AIF are common practice. We recently transferred this approach to small animal dynamic contrast-enhanced (DCE)-MRI and demonstrated successful recordings of the AIF in extracorporeally perfused blood reservoirs placed inside the MRI field of view ¹. However, the introduced approach left open questions regarding the (a) ideal vascular access ways, (b) quantitative precision, (c) ideal dispersion correction and (d) the applicability with compressed sensing sequences. In the current study, we address these open questions by (a) comparing vascular access through the femoral artery and the carotid arteries, (b) validating the quantitative precision with simultaneous recordings of extracorporeal contrast agent radiotracer analogs and with inductively coupled plasma mass spectrometry (ICP-MS), (c) comparing two different dispersion correction methods and (d) applying state of the art compressed sensing and parallel imaging sequences for DCE and T1-mapping.

Methods

Overall 23 female NMR nu-/nu- mice were measured in a 9.4 T Bruker BioSpec small animal MRI. All mice had intracranial U87 glioblastoma tumors. Either 100 µl of 35 mM Gd-DO3A-butrol (Gadovist) or 100 µl Gd-DTPA (Magnevist) were mixed with their radioactive analogs [⁶⁸Ga]-Ga-DO3A-butrol (5.74 ± 2.96 Mbq) or [^99m Tc]-Tc-DTPA (23.78 ± 4.08 Mbq) and simultaneously injected. An extracorporeal circulation was applied either from the femoral artery to the tail vein (n = 13) or from the carotid bifurcation to the jugular vein (n = 10). The circulations featured MR measurement reservoirs with a volumetric distance of 16.1 µl from the femoral artery and a negligible distance from the carotid artery (1.6 µl). Further downstream, blood radioactivity was assessed in a MR compatible measuring unit (Swisstrace, Twilite) and additionaly, laser doppler flow measurements were performed. A Golden-angle Radial Sparse Parallel (GRASP) sequence (approximately 0.195 x 0.195 x 0.195 mm³, 5 s/frame, TR/TE 4/0.011 ms, FA 13.4°)², combined with a compressed sensing MP2RAGE sequence for T1 mapping (approximately 0.195 x 0.195 x 0.195 mm³, TI₁/TI₂/MP2RAGE_TR = 800 ms/2200 ms/6250 ms, TE/TR = 3.253/7 ms, FA 7°)³, was used for DCE-MRI (Figure 1). Contrast agent concentrations were calculated as previously described ¹ and were based on Relaxation Rates of r₁ = 4.1 mM⁻¹s⁻¹ for Gadovist and r₁ = 2.94 mM⁻¹s⁻¹ for Magnevist. T1 values were individually measured with cs-MP2RAGE for extracorporeal blood. Literature value of T1 = 2400 ms was taken for intracorporeal blood. Radioactive tracer concentrations were determined based on a separately measured sensitivity factor. Deconvolution of the resulting AIFs from the proximal reservoir were computed by performing two different approaches, either a parametric model using a fitting routine a non-parametric model, using a regularized iterative least square method. Both were depending on the estimation of the dispersion kernel, derived by the convolution between the AIF extracted from the distal reservoir and a single gamma variate.

Results:

Vascular access was experimentally feasible for the femoral artery and vessels of the neck. Whereas dispersion correction was necessary for the femoral artery setup, curves were not significantly dispersed for the carotid artery setup, where curve geometry closely resembled intracorporeal measurements form retroocular vessels (Figures 2-4). Both access ways showed a high degree of correspondence between extracorporeally determined contrast agent concentrations and simultaneously acquired recordings from radiotracer analogs (Figure 2-4) for both contrast agent / radiotracer analog pairs. Similarly for 10 mice measured with the femoral artery setup, equal good correspondence from blood samples, between MRI and ICP-MS determined Gadolinium concentrations was observed (mean deviation 8.7 ± 28 %). In contrast, correspondence of intracorporeally derived AIFs from retroocular vessels was highly variable and overall inferior to extracorporeal measurements. Dispersion effects in the femoral artery shunt were successfully corrected using kernels based on single gamma variate functions, resulting in AIFs resembling the carotid shunt where dispersion seemed negligible (Figure 3-4). Both dispersion correction strategies demonstrated exceedingly realistic AIFs with kernels resulting in excellent fits e.g., R² > 0.9997.

Discussion

Shunt flows varied across different animals but were rather constant in individual measurements. The resulting impact of varying shunt flow on deconvolution strategies isn´t fully solved yet. Therefor it has to be determined in the future, unclear which vascular access way is more robust and precise in subsequent PKM.

Conclusion

Independent of the vascular access way, extracorporeal derived AIFs seem to be a promising and feasible strategy for determination of AIFs in mice with high quantitative precision. Consistency of the two dispersion models in individual mice prove the use of gamma variate functions as dispersion kernels. The use of novel rapid imaging techniques in combination with extracorporeal AIF bears a high potential to set the basis for precise quantitative dual PET/MRI PKM in mice.

Acknowledgements

The project was supported by the SFB CRC 1450 and the DFG-NAMT application 446292030 of Dr. med. Philipp Backhaus. The project critically relied on support of technical staff, especially Roman Priebe, Dirk Reinhardt, Nina Kreienkamp and Irmgard Hoppe.