Myelin water fraction (MWF) is a good biomarker for myelin content. Traditional methods for acquiring MWF maps require long scan times. Recent work has estimated MWF from faster steady-state scans. In this work, we propose to acquire MWF maps from an optimized set of small-tip fast recovery (STFR) scans that can exploit resonance frequency differences between myelin water and the slow-relaxing water compartment.
Myelin water fraction (MWF), the proportion of MR signal in a given voxel that originates in water bound within the myelin sheath, is a specific biomarker for myelin content. Such a biomarker is desirable for tracking the onset and progression of demyelinating diseases such as multiple sclerosis. A common way to estimate MWF is from a multi-echo spin echo (MESE) pulse sequence, which is time-consuming.1 Recent work has estimated MWF from fast steady-state sequences.2 In this abstract, we propose estimating MWF from an optimized set of small-tip fast recovery (STFR) scans that can exploit resonance frequency differences between myelin water and the slow-relaxing water compartment. Simulation results illustrate how well STFR scans can estimate MWF.
The STFR pulse sequence3 consists of a tip-down RF pulse, signal readout, a tip-up RF pulse, and a spoiler gradient. The transverse signal generated by a single STFR scan at a particular voxel (for a single compartment) is
MT(M0,T1,T2,Δω,κ,Tfree,Tg,α,β,ϕ)=M0sin(κ⋅α)[e−Tg/T1(1−e−Tfree/T1)cos(κ⋅β)+(1−e−Tg/T1)]1−e−Tg/T1e−Tfree/T2sin(κ⋅α)sin(κ⋅β)cos(Δω⋅Tfree−ϕ)−e−Tg/T1e−Tfree/T1cos(κ⋅α)cos(κ⋅β),
where M0 is proton density, T1 is the spin-lattice relaxation time constant, T2 is the spin-spin relaxation time constant, Δω is off-resonance frequency, κ is a flip angle scaling factor (to account for imperfect transmit fields), Tfree is the time between the tip-down and tip-up pulses, Tg is the duration of the spoiler gradient, α is the prescribed tip-down flip angle, β is the prescribed tip-up flip angle, and ϕ is the phase of the tip-up pulse. Note that M0, T1, T2, Δω, and κ vary from voxel to voxel, whereas Tfree, Tg, α, β, and ϕ are scan parameters that are prescribed over the whole imaging volume.
We consider two non-exchanging intra-voxel water compartments: a fast-relaxing compartment with relaxation time constants T1,f and T2,f, and a slow-relaxing compartment with relaxation time constants T1,s and T2,s. We assume the fast-relaxing compartment experiences an additional off-resonance shift Δωf.4 The signal from a given voxel is a weighted sum of the signal that arises from the fast-relaxing and slow-relaxing compartments, where the weights are ff and 1−ff, respectively, and ff denotes the fraction of the signal arising from the fast-relaxing compartment. We estimate the MWF ff for each voxel from multiple STFR scans.
We optimized a set of 9 STFR scans to maximize the precision of estimates of ff. We minimized the expected Cramer-Rao Bound (CRB) of estimates of ff.5 We fixed Tfree to 8.0 ms and Tg to 1.5 ms and optimized α, β, and ϕ. For Δω and κ we used separately acquired B0 and B1 maps, respectively. Table 1 lists the optimized scan parameters.
Using the optimized scan parameters, we simulated the 9 STFR scans using a slice of the BrainWeb phantom.6 For white matter, we assigned M0=0.77, ff=0.15, T1,f=T1,s=832 ms, T2,f=20 ms, T2,s=80 ms, and Δωf=17 Hz; and for gray matter, we assigned M0=0.86, ff=0.03, T1,f=T1,s=1331 ms, T2,f=20 ms, T2,s=80 ms, and Δωf=0.2 We generated κ to vary from 0.8 to 1.2 (i.e., 20% flip angle variation), and Δω to vary from -20 to 20 Hz. We added complex Gaussian noise to produce images with SNR ranging from 89-244 in white matter and 64-236 in gray matter, where SNR is defined as \mathrm{SNR}\left(\mathbf{y},\boldsymbol{\epsilon}\right) \triangleq \frac{\|\mathbf{y}\|_2}{\|\boldsymbol{\epsilon}\|_2}, where \mathbf{y} is the noiseless data within a region of interest (ROI), and \boldsymbol{\epsilon} is the noise added to the ROI. We estimated f_\mathrm{f} from the STFR images using kernel machine learning.7
[1] A. Mackay, K. Whittall, J. Adler, D. Li, D. Paty, and D. Graeb. In vivo visualization of myelin water in brain by magnetic resonance. Mag. Res. Med., 31(6):673–7, June 1994.
[2] G. Nataraj, J-F. Nielsen, M. Gao, J. A. Fessler. Fast, precise myelin water quantification using DESS MRI and kernel learning. arXiv:1809.08908v1 [physics.med-ph].
[3] J-F. Nielsen, D. Yoon, D. Noll. Small‐tip fast recovery imaging using non‐slice‐selective tailored tip‐up pulses and radiofrequency‐spoiling. Mag. Res. Med., 69(3):657-66, March 2013.
[4] K. Miller, S. Smith, P. Jezzard. Asymmetries of the balanced SSFP profile. Part II: white matter. Mag. Res. Med., 63(2):396-406, February 2010.
[5] G. Nataraj, J-F. Nielsen, and J. A. Fessler. Optimizing MR scan design for model-based T1, T2 estimation from steady-state sequences. IEEE Trans. Med. Imag., 36(2):467–77, February 2017.
[6] D. L. Collins, A. P. Zijdenbos, V. Kollokian, J. G. Sled, N. J. Kabani, C. J. Holmes, and A. C. Evans. Design and construction of a realistic digital brain phantom. IEEE Trans. Med. Imag., 17(3):463–8, June 1998.
[7] G. Nataraj, J-F. Nielsen, C. D. Scott, and J. A. Fessler. Dictionary-free MRI PERK: Parameter estimation via regression with kernels. IEEE Trans. Med. Imag., 37(9):2103–14, September 2018.