5512 Extravascular extracellular space fraction measurement by DSC-MRI : a theoretical study
Yikang Liu,Weiping Ding,Bensheng Qiu
2013-01-01
Abstract:Introduction: The unclear T1 and T2 effect dominance resulting from the contrast agent (CA) leakage in dynamic susceptibility contrast (DSC)-MRI weakens linearity between ΔR2* and CA concentration. But endeavor has been made to measure relative cerebral blood volume (rCBV) and blood-brain barrier (BBB) permeability by DSC-MRI, such as Weisskoff model and its modified forms [1]–[4]. However, no one, to our knowledge, has discussed the relation between extravascular extracellular space (EES) fraction Ve and parameters deduced from Weisskoff model. In this study, our results from simulation demonstrated that the ratio of K1 and K2 in Weisskoff model was linear to Ve, thus illustrating a potential approach to measure relative Ve by DSC-MRI. Methods: A simulation tool was developed in Matlab (The Mathworks Inc.) and COMSOL Multiphysics (CnTech Co.), based on approach proposed by Pannetier at al (2013)[5], which comprised blood flow, CA leakage, CA diffusion in EES, intrinsic and CA-induced R1 and R2 relaxations, magnetic field perturbations, and the diffusion of the water protons. A voxel was modeled as a 100*100μm surface with cells and vessels modeled as circles randomly distributed. The perturbation of the magnetic field was averaged over 3 orthogonal orientations [5]. The diameter of vessels was set as 4μm according to that of microvessels in grey matter [6] and distribution of cell diameters were set according to axon diameter distribution of optical nerves [7]. Arterial input function (AIF) was acquired from in vivo DSC data fitted to gammavariate function and cerebral blood fraction Vb was 3.8%[5]. Transfer coefficient Kpe of vascular wall was set as 0, 0.001, 0.009, 0.016, 0.036, and 0.064, and Ve was set as 0.2 to 0.8 with increment of 0.1. Both included the range reported in tumors[5][8]. And a preload of 0.01 mM was implemented. In addition, single-shot gradient echo sequence was used for simulation with TR = 1500ms, TE = 10ms, and flip angle = 90 degrees, lasting for 100s. Susceptibility maps were converted to magnetic field perturbation maps by the Fourier transform approach and MR signals were simulated by Bloch equations under 4.7 T [5]. Five simulations were conducted for each combination of the parameters. TheΔR2* curves of tissue and AIF were generated from MR signals byΔR2* = ln(S(0)/S(t))/TE, with Vb considered for AIF. Then, the curves were fitted into the Weisskoff model through Levenberg-Marquardt algorithm, and linear relationship between Ve and K1/K2 was evaluated by Pearson's linear correlation coefficient. Results: Fig. 1 illustrates the relationship between K1/K2 and Ve for Kpe = 0.001, 0.009, 0.016, 0.036, and 0.064 respectively. Pearson’s coefficients and parameters of linear regression equation K1/K2 = m*Ve+b for each circumstance are displayed in Table 1, which manifest linear correlation between K1/K2 and Ve and their stable relationship. Fig. 2 shows the magnetic field perturbation ΔB maps of Ve = 0.2 and Ve = 0.8 when B0 parallels and orthogonal to the surface respectively. Discussion: T2 effect is on account of dephasing of MR spins caused by magnetic field perturbation. As shown in Fig. 2, higher Ve leads to larger susceptibility interface area and thus more significant T2 effect. On the other hand, higher Ve results in a slower extravasation process and thus undermines T1 effect. As a result, the absolute value of K1/K2, which is positively correlated to T2 effect and negatively correlated to T1 effect, is lower with higher Ve. Furthermore, the relationship between Ve and K1/K2 is almost invariant for different permeability, as shown in Fig.1 and Table 1, the reason of which may be as follows. First, the CA in EES causes T1 and T2 effects at the same time, which contradict with each other, then the amount of CA may have weak relation with MR signal. Second, the integral of tissue response function in Tofts model[9] Kpe*e on timespan of 0 to 100s varies little with a certain range of Kpe, for example, 0.907 for Kpe =0.025, 0.969 for 0.05, and 0.951 for 0.1, effect of Kpe therefore diminished. Conclusion: The stable relationship between K1/K2 and Ve under conditions of different permeability and cell/vessel distribution has been manifested by numeric simulation. As a consequence, Weisskoff model could be utilized for generating relative Ve map by DSC-MRI and clinical data is expected to further prove the theory. References: [1] R. Leigh at al, PLoS One, vol. 7, no. 12, 2012. [2] J. L. Boxerman at al, AJNR Am J Neuroradiol, vol. 27, no. 4, 2006. [3]A. Bjornerud at al J. Cereb. Blood Flow Metab., vol. 31, no. 10, 2011. [4] H. L. Liu at al, Med Phys, vol. 38, no. 2, 2011. [5] N. A. Pannetier at al, PLoS One, vol. 8, no. 3, 2013. [6] A. D. and D. Boas, Neurovascular Imaging. Frontiers E-books. [7] Y. Assaf at al, Magn. Reson. Med., vol. 59, no. 6, 2008. [8] A. Singh at al, J Magn Reson Imaging, vol. 29, no. 1, 2009. [9] P. S. Tofts at al, J Magn Reson Imaging, vol. 10, no. 3, 1999 Kpe R P m b 0.001 0.986 4e-5 83.59 -81.25 0.009 0.988 3e-5 83.04 -81.15 0.016 0.987 4e-5 83.43 -81.54 0.036 0.985 6e-5 84.87 -81.85 0.064 0.988 3e-5 84.05 -81.68 Table 1. Relationship between Ve and K1/K2. R and P are Pearson’s coefficients; m and b are parameters in linear regression equation.