An APTw/CEST-MRI technique to obtain water-only Z-spectra in the presence of fat is described. Water-fat separation in CEST is complicated by partial saturation of spectral fat components. Here, the saturation frequency-dependent water-fat phases are calculated using a multi-peak saturation model. A Z-spectrum acquisition with a single echo-shift is combined with a reference acquisition (S0) using 3 echo-shifts for water-fraction and B0-mapping. A B0-corrected, water-only Z-spectrum is obtained by complex rotations and weighted subtraction according to the water-fat phase. Volunteer examinations at 3T are shown (breast and abdominal). The single echo-shift technique offers a time-efficient means for water-fat separation in APTw/CEST-MRI.
For the Z-spectrum, complex images S[ω]={SRe,SIm} are acquired with a single echo shift (ES). As reference, S0 (-1560ppm) is measured with 3 different ES for full water(W0)/fat/B0 assessment using a 3-point Dixon technique4,5. A multi-peak fat saturation model Fsat[ω], assuming a saturation linewidth of Δ= 1ppm, is employed to calculate the water-fat phase angle α[ω], which varies with ω because the fat spectrum is partly saturated by CEST contrast preparation (see Fig.1):
$$$F_{sat}[ω]=\sum_{k=1}^{N}p_k(1−L_k[ω])e^{i α_k}$$$
$$$L_k[ω]=\frac{Δ^2}{4(ω−ω_{F_k})^2+Δ^2}$$$
ωFk and pk denote the relative frequency and area of the individual peaks of the fat spectrum. α[ω] is then derived from the phase angle of Fsat[ω]:
$$$α[ω]=atan(Im[F_{sat}[ω]],Re[F_{sat}[ω]])$$$
A complex rotation on {SRe,SIm} with angle ε aligns the water signal with the real axis in each voxel and for each ω:
$$$ε=α[ω_0]−\phi_0−arcsin[\frac{W_0}{S_0}⋅sin(\pi−α[ω_0])]$$$
$$$\phi_0=atan(S_{Im}[ω_0],S_{Re}[ω_0])$$$
On the resulting complex data, a B0 correction is performed by Lagrange interpolation, shifting the spectral center according to B0. The water-only Z-spectrum is finally obtained via a combination of the resulting complex signals {SRe',SIm'}, similar to a single-point Dixon technique6:
$$$Re_{LC}[ω]=\frac{1}{S_0}[S_{Re}' − S_{Im}'⋅tan(\frac{\pi}{2}−α[ω])]$$$
This constitutes a subtraction of the projection of the lipid signal onto the real axis. The projection depends on the phase angle α[ω]. As an example for Z-spectral analysis, asymmetry analysis can be performed on ReLC[ω], providing water-only MTRasym;LC. Volunteer examinations were performed on 3T MRI systems with dual-channel body coil RF transmission (Achieva TX, Philips Healthcare, Netherlands) in the breast, using a 16-channel bilateral breast coil, as well as in the lower abdomen, using a 32-channel torso coil. RF saturation consisted of 40×50ms pulse elements alternatingly transmitted via the RF channels at 100% duty cycle, with a total duration of 2s and a B1,rms of 1.2μT (breast)/ 2μT (abdomen). For the breast examinations, a triple gradient-echo Dixon sequence was used with TR/TE1/∆TE = 5.2/1.58/1.0ms, flip angle (FA) = 10° and centric k-space ordering (low-high) at 34 different saturation frequency offsets from -6 ppm to 6 ppm and S0. For the single-echo reconstruction, only TE1 was used, except for the reference S0. In the abdominal example, Z-spectra were acquired with a single-shot fast spin echo (FSE) readout using TR/TE=6s/5.9ms, voxel size (1.8)2x5mm3, echo shift ES1=-0.52ms at 43 different saturation frequency offsets from -9.2 ppm to 9.2ppm (step 0.44ppm) and (S0). S0 was acquired with 3 different ES1,2,3=-0.52/0.0/+0.52ms.