Myelin is essential for proper functioning of the central nervous system. The ability to measure myelin density noninvasively would have a major impact on diagnosis and evaluation of diseases that are responsible for much of CNS morbidity. The NMR spectral properties of myelin have previously been shown to be consistent with a lamellar liquid crystal yielding a Superlorentzian line shape with broad tails extending to about ±20 kHz, resulting in lifetimes of the transverse magnetization on the order of tens of microseconds [1]. Here we examined the potential of 3D UTE and ZTE solid-state MRI with long-T₂ suppression to quantify myelin in native lamb spinal cord using reference samples containing various mass fractions of reconstituted myelin at 9.4T.

Sample Preparation: Myelin was extracted from bovine spinal cord by a sucrose gradient technique[2] and suspended in 99.9% D₂O to achieve concentrations of 6%, 8%, 10%, 12% and 14%, contained in 5mm NMR tubes serving as reference samples. A 36-mm segment of cervical spinal cord was dissected from sheep and stored in H₂O-phosphate buffered saline until use.

Imaging Experiments: All experiments were performed at 9.4T (Bruker Avance) with maximum gradient amplitude of 1,000 mT/m. Previous work showed the UTE signal amplitude to be strongly correlated with the mass fraction of myelin suspended in D₂O(R²=0.98) [1]. Here we investigated the potential of both UTE and ZTE imaging for actual quantification of myelin in native neural tissue with an experimental set-up similar to that used in [1]. Inclusion of ZTE was motivated by earlier observations of its superior SNR relative to UTE [3]. A commercial ZTE sequence was used to image the reference samples (TR=2ms, FA=4.1o, 2µs pulse duration, 3.2µs dwell time, 51,500 half-projections, 250×250×750µm³ resolution, scan time=1.7 min). The same samples were also scanned with a long-T₂ suppressed IR-ZTE derived from the commercial sequence after insertion of an adiabatic inversion pulse preceding excitation. One half-radial spoke was scanned following each inversion (TR=200ms and TI=70ms resulting in 3 hours for 51,500 half-projections). To increase SNR, the excitation flip angle was set to 55^o, which, however, prolonged pulse duration to 25μs and dwell time to 10μs. Both ZTE and IR-ZTE image signals were regressed against myelin fraction of the reference samples.

To quantify myelin in situ in the native lamb spinal cord, four pulse sequences (ZTE, IR-ZTE, UTE and IR-UTE) were run along with the myelin reference samples using the protocols described above except that excitation the flip angle of IR-UTE was set to 90^o to maximize the signal and spinal cord gray and white matter (GM, WM) myelin fraction was estimated from the calibration equation derived from the association between MR signals of the reference samples and their myelin fractions.

The ZTE and IR-ZTE images of the reference samples are shown in Figs.1a and b. Both ZTE and IR-ZTE signals are linearly correlated with the myelin fraction as expected (R²=0.99 and R²=0.97, respectively, Figs.1c and d). SNR was ~25 in the ZTE image of 14% myelin/D₂O suspension. However, the IR-ZTE image suffers from inferior SNR (<5) and blurring artifacts due to the much prolonged pulse duration and sampling window (as a consequence of peak power limitation since the flip angle can only be increased by prolonging pulse duration) .

Figs.2a-c show the spinal cord images acquired with ZTE, UTE and IR-UTE. SNR of the ZTE, UTE and IR-UTE images of the 14% myelin/D2O suspension was 25, 8 and 15, respectively. GM-WM contrast is substantially enhanced in IR-UTE relative to UTE due to the higher myelin content in WM and suppression of long-T₂ tissue water. From the linear relationship between IR-UTE signals of reference samples and their myelin fractions (R²=0.97, Fig. 2d), an apparent WM myelin fraction of 22.3% is obtained. However, since reconstituted myelin does not contain proteins (~30%), the estimated myelin lipid fraction is ~16%. Further possible sources of error are other cell membrane short-T₂ tissue constituents as well as incomplete suppression of long-T₂ signal.

The potential of UTE and ZTE MRI for direct myelin quantification was evaluated on myelin extract and ex vivo lamb spinal cord. In the present implementation, IR-preparation suppresses the long-T₂ signal close to the background noise level and IR-UTE achieves the most promising results. Although superior SNR is achieved with ZTE, long-T₂ suppressed IR-ZTE suffers from poor SNR inadequate for quantification due to excitation flip angle limitations. Further work will focus on improvements on long-T₂ suppressed ZTE imaging for directed myelin quantification, e.g. by means of multiple readouts following each inversion, and maximizing flip angle with phase-modulated pulses[4].