Introduction:

The MR vascular fingerprinting (MRvF) approach¹ extends the concept of MR fingerprinting² to the study of microvascular properties and functions. The MR signal evolution (Free Induction Decay + Spin echo=fingerprint) is sampled in each voxel and compared to a dictionary of curves obtained using advanced numerical simulations of the same experiment. The results are high-resolution parametric maps representing blood volume fraction (CBV), blood vessels radius (R), and blood oxygenation (SO₂). Encouraging results have been obtained in healthy human volunteers¹ as well as in stroke and tumor models in rats³. However, it has been suggested that the method has a low sensitivity to blood oxygenation measurements^1,4. In this study, we tested the MRvF approach in healthy volunteers while breathing different gas mixtures (Hyperoxia (100%O₂), Normoxia (21%O₂), hypoxia (14%O₂)) and examined the results when different types of fingerprints are considered.

Materials and Methods:

The local IRB committee approved all studies. Experiments were performed at 3T (GE Healthcare Systems, Waukesha, WI) with an 8-channel GE receive-only head coil. The MR acquisition protocol included: (1) a blood T2 mapping sequence based on the TRUST approach⁵ for independent blood oxygenation measurements (acq matrix=64x64, FOV=22x22cm², NA=4, TR=10s, TEs=20-40-60-100ms, Acq time=264s). (2) A GESFIDE sequence for fingerprinting (TR=2000ms, 40echoes, SE=100ms, FOV=20*20cm², ST=1.5mm, 128*128, 12slices, Tacq=4min). 10 volunteers were scanned while breathing different gas mixtures (Hyperoxia (100%O₂), Normoxia (21%O₂), hypoxia (14%O₂)). Each epoch lasted 6 minutes and acquisitions started after 1 min to ensure stable gas environment and subject physiology. The protocol was repeated after intravenous injection of Feraheme (off label contrast agent, 7 mg Fe/kg, AMAG Pharmaceuticals Inc., Cambridge, MA, USA) and the ratio of the GESFIDE acquisitions pre and post contrast was initially taken as ‘vascular fingerprint’¹. Simulations and post processing were performed as previously described¹.

Results:

Physiological monitoring and TRUST measurements (Fig.1) indicated global changes of blood oxygenation during the gas challenges. MRvF parametric maps obtained in one volunteer during the 3 challenges are presented in Fig.2a-c. CBV maps show a good contrast between GM and WM, while Vessel Radius and Blood oxygen saturation maps are more homogeneous. The numerical values obtained during normoxia and averaged over the volunteers are consistent with previous reports. However, no changes were observed in SO₂ estimates between the gas challenges. The accuracy of the fingerprinting approach depends on the ability to separate the fingerprints signals in the dictionary. In order to study the sensitivity of MRvF to SO₂, we examined the dictionaries (pre-contrast, post-contrast, ratio pre/post, Fig.3a-c) when CBV and R are fixed (Fig.3d-f). It can be observed that the maximum sensitivity to SO₂ appears in the pre-contrast dictionary (green arrow) while only a portion of the ratio dictionary allows for curves discrimination. These findings are summarized in Fig.3g-i where the standard deviation of the curves is computed at each echo time. Figure 4 extends the previous results for all parameters. It confirms that the maximum sensitivity to SO₂ is obtained in the pre-contrast dictionary (although CBV also influences the results) and shows that the method is mostly sensitive to CBV changes (especially using the post-contrast dictionary). R measurements should be extracted from the ratio dictionary. Given these observations, we designed a new fingerprint that concatenates the pre, post and ratio signals. Figure 5 shows the new results after using this updated dictionary. The CBV and R results are similar to those initially obtained but the SO₂ estimates now show significant differences between hypoxic and hyperoxic measurements.

Conclusion:

This study shows that MRvF maps can be obtained in healthy volunteers while modifying the inspired oxygen content. Results confirm that the original MRvF fingperprint pattern has a low sensitivity for SO₂ measurements. By analyzing the properties of the dictionaries, we show that it is possible to design a new fingerprint pattern with higher SO₂ sensitivity which seems promising for non-invasive detection of hypoxic tissues in several pathologies.

Acknowledgements

Supported in part by (NIH 5R01NS066506, NIH 2RO1NS047607, NCRR 5P41RR09784, R21-NS087491).

References

[1] Christen et.al, Neuroimage, 2014 [2] Ma et.al, Nature, 2013 [3] Lemasson et.al, Scientific Reports 2016 [4] Pouliot et.al, Neuroimage 2017 [5] Lu and Ge, MRM, 2008.