Introduction

The oxygen extraction fraction (OEF) calculated from venous oxygenation (Y_v) is an important hemodynamic indicator of brain health and function. T₂-oximetry based Y_v mapping methods such as QUIXOTIC^1,2 and VSEAN^3,4 currently suffer from low SNR of available venous blood signal in each voxel. Recently, a venous CBV method⁵ by utilizing Fourier-transform based velocity-selective inversion (FT-VSI) and saturation (FT-VSS) pulse trains for arterial-nulling and tissue suppression was introduced, in order to achieve higher SNR and insensitivity to arterial transit delay and CSF contamination⁶. Here we combine this venous isolation preparation with the multi-echo acquisition to estimate Y_v.

Methods

The FT-VS pulse trains (Fig. 1a) are composed of nine excitation pulses, interleaved with paired and phase-cycled refocusing pulses surrounded by gradients with alternating polarity⁷. The normalized Mz-velocity responses of FT-VSS and FT-VSI are displayed in Figs. 1b,c. The non-selective inversion (NSI) pulse applied immediately after FT-VSI flips the VSI profile such that the flowing spins within the passing band get inverted and the static spins within the inversion band are restored (Fig. 1c, red). Fig. 1d outlines the sequence diagram for estimating Y_v. The ideal evolution of the microvascular signal through the preparation phase of this sequence is illustrated in Fig. 2. While the spins flowing above the V_CUTOFF would be inverted and then get nulled at the end of TI, the spins moving below the V_CUTOFF such as those within the capillaries, would largely flow out into the venules during TI with the preserved signal magnetization. After the subtraction of the results of FT-VSS control and labeling scans (V_CUTOFF=0.7cm/s), only the spins within the venular compartment that accelerated above the V_CUTOFF will be retained for imaging.

Two healthy volunteers provided informed consent to the institutionally approved protocol. Experiments were conducted on using a 3T Philips Ingenia scanner with a 32-channel head coil reception. The imaging parameters were: 8mm slice thickness, FOV= 252×252mm² with the acquisition resolution of 4.0×4.3mm², SENSE factor of 2.3 for phase-encoding and TR of 4.5s. Two shots were used with an EPI factor of 13 to fill the k space along the phase-encoding direction with a phase-cycling of [0, π] applied on the excitation pulse. Twenty echoes were acquired with an echo spacing step of 14ms. Hyperbolic tangent pulses (4ms, frequency sweep of 11500Hz) were used in the acquisition for immunity to B₀/B₁ inhomogeneities. Eight dynamic scans were obtained with a total scan time of 5.2min. With the same FOV and resolution, double inversion recovery (DIR) was applied for visualizing gray matter (GM) and white matter (WM) only. CSF T₂ was also measured by combining a 600ms T2prep, in order to suppress tissue signal⁸, and 32 echoes after the excitation with an echo spacing of 40ms. As a comparison, blood T₁ and T₂ were measured at the internal jugular vein (IJV)^9,10 to quantify Hct and global Y_v^11,12.

To remove the effect of the residual CSF signal¹³, the image of the last echo (280ms), which has only CSF signal remaining, was subtracted from the image of the nth echo using the equation¹³
$$$SI_{n,corrected} = SI_{n} - e^{\frac{(20-n)\cdot14ms}{T_{2,CSF}}} \times SI_{20}$$$
The corrected images were resized to a 32×32 matrix (8mm resolution), and then smoothed with a 9mm FWHM Gaussian kernel¹. The T_2v map was obtained by fitting all 10 even echoes voxel-by-voxel using the mono-exponential decay function:
$$$S_{v} = S_{0} \times e^{-TE/T_{2v}}$$$
The obtained T_2v map was converted to the Y_v map using the previously developed T₂-Y calibration model¹¹ based on the Hct calculated from the measured blood T_1v¹⁴. The standard error of the voxel-wise T₂ fitting (T_2error) was recorded as well, indicating the uncertainty of the fit. The two-parameter fitting also determined the scaling factor S₀, which is proportional to the venous CBV. Lastly, the temporal SNR (tSNR) of the second echo through the remaining dynamics was also computed for each voxel.

Results

Fig. 3 shows the DIR GM, S₀, T_2v and Y_v maps for the two subjects. As expected, based on CBV, the calculated S₀ maps show higher signal in GM than in WM, which is consistent with the venous CBV contrast. Conversely, the fitted T_2v maps and the corresponding Y_v maps delineate a more uniform contrast between GM and WM (Table 1), as expected for brain OEF. The averaged Y_v for the two subjects (Table 1) are 0.56±0.07 and 0.56±0.09, very close to their global Y_v of 0.54 and 0.61, respectively.

The maps of tSNR and T_2error are shown in Fig. 4. Note that the voxels with lower tSNR and higher T_2error values are mainly in the WM area in line with the lower blood volume. The T_2error values are negatively correlated with tSNR (p<0.001). The T_2v fitting plots from representative voxels with different tSNR are also compared.

Conclusion

Employing advanced velocity-selective pulse trains, a new venous oxygenation mapping technique is proposed with high venous signal and minimal sensitivity to arterial transit delay and CSF contamination. The measured venous oxygenation values are similar between GM and WM and comparable to the values measured globally.

Acknowledgements

We thank Joseph Gillens, Terri Brawner, Ivanan Kusevic and Kathleen Kahl for their experimental assistance, and the funding support from NIH (P41 EB015909, NIH: K25 HL145129, NIH S10 OD021648, R01 HL138182, NIH: R01 HL144751)