Introduction

The mechanisms underlying CSF production, glymphatic clearance, and CSF reabsorption have become a focus of interest for the study of Alzheimer’s disease and other pathologies. A key process in CSF dynamics is the exchange of water across compartments, from tissue and blood into the CSF, subarachnoid, and paravascular spaces. T₂ selective preparation may be a useful approach to MRI characterization of this exchange¹.

Theory

Our approach to T₂ selective labeling employed a second image with T₂ saturation applied later as a control, figure 1. For the labeled sequence, a T₂ selective saturation is applied before a mixing time T_mix. For the control sequence, The T₂ selective saturation is applied after T_mix. In the absence of exchange during the mixing time, the magnetizations after the label and control sequences are given by

$$$M_{2lbl}={\alpha}M_1(e^{({-T_{mix}/T_1})})+R$$$
$$$M_{2ctl}={\alpha}M_1(e^{({-T_{mix}/T_1})})+{\alpha}R$$$

Where alpha is the saturation factor for the T₂ saturation and R is the recovered magnetization during the mixing period. The difference between the two ending magnetizations is nonzero only because of the recover term R.

$$$M_{2lbl} - M_{2ctl} =(1-\alpha)R= (1-\alpha)(1-e^{({-T_{mix}/T_1})})$$$

If we add n inversion pulses during the mixing time, the M₁ terms above are multiplied by a power of the inversion efficiency. But, if the timing of the inversion(s) is optimized, one can make R close to zero². For 3 or more inversions, R can be reduced to less than 1% for T₁’s from pure water to fat.

$$$M_{2lbl} - M_{2ctl}= \alpha\beta^nM_1e^{({-T_{mix}/T_1})}-\alpha\beta^nM_1e^{({-T_{mix}/T_1})}+(1-\alpha)R$$$

Because this subtraction removes direct effects of labeling on exchanging spins, any difference between label and control should reflect exchange during the mixing time such that T₂ and/or T₁ are not constant.

Methods

The T₂ selective exchange sequences were implemented on a GE Signa Premier XT scanner before a 3DFSE (RARE) sequence. T₂ preparation was achieved with a 200ms TE BIR8 adiabatic sequence³, and 4 tanh adiabatic inversion pulses were applied at optimized times to minimize recovered magnetization. The preparations were preceded by nonselective saturation 5 s before imaging and a T₂ selective inversion recovery optimized to nearly null CSF M₁. Following the preparations, 200ms were allowed to allow some recovery of tissue magnetization and 3 fat saturations pulses were applied immediately before imaging. A TR of 10s, 2x2 parallel imaging acceleration, an asymptotic 70° flip angle train with echo spacing of 3.3ms, and centric phase ordering with TE controlled by skipping echoes prior to acquisition were selected. Acquisition of the label and control images and an unprepared reference image required 5min 20s.
Aftertesting in MnCl phantoms with a range of T₁’s and T₂’s that showed subtraction errors of less than 0.1%, images were acquired in 3 healthy volunteers following an IRB protocol and written informed consent. Images were acquired for TE’s of 106.5, 213.0 and 319.5ms for T_mix of 2s and 1.5s and for the 2 longer TE’s at T_mix of 1s. DICOM images were processed in MATLAB and SPM. Following gaussian smoothing to 3x3x3 mm resolution, label and control images were subtracted and divided by the signal in the center of the ventricles on the reference image.

Results

All images showed elevated signal surrounding the choroid plexus and distributed throughout cortical and brain stem regions, figure 2. Negative white matter signal was noticeably present on the TE 106.5 ms images but the effected faded by the 213 ms images and was negligible in the 319.5 ms images. This effect likely reflects incomplete suppression of recovered magnetization due to a very short T₁ component in white matter⁴. Exchange signal in choroid plexus and near cortex appeared to increase slightly with TE, consistent with reduced partial volume of blurred negative white matter signal. Though the exchange signal was present only in regions known to contain CSF, this signal could not simply be a systematic error in CSF, since the spatial variation of intensity was very different from the unsubtracted label or control images and the reference images. 3D images averaged across subjects show the whole brain distribution of the exchange signal, figure 3. The spatial distribution of exchange signal was consistent across subjects with the highest signal around the choroid plexus of the lateral ventricles. Signal was also prominent in the fourth ventricle and around the cerebellar vermis. Noticeable exchange can be seen surrounding the cerebellar and cerebral cortices.

Conclusions

An approach for sensitive measurement of exchange from short to long T₂ compartments can be used to assess water exchange from tissue and blood to CSF. This method may be used to help understand and diagnose disorders of CSF production and the glymphatic clearance system.

Acknowledgements

No acknowledgement found.