The purpose of this work was to characterize and image ultrashort-T2 components in the brain, which have been presumed to originate from protons in the myelin phospholipid membranes.

We used a 3D UTE pulse sequence with a non-selective, hard pulse excitation and radial k-space readouts, with 16 TEs interleaved within a single acquisition (Fig. 1). TEs ranged from 100 $$$\mu s$$$ up to 4 ms in order to focus on the ultrashort-T2 components. The scan time was 15 minutes for 2.5 mm isotropic resolution and whole brain coverage (FOV = 24 cm), using TR = 10 ms, and a 512 $$$\mu s$$$ readout duration. All scans we performed on a 7 T MRI system (GE Healthcare) using a head-only quadrature coil to transmit and a 32-channel phased-array head coil (Nova Medical) to receive signal.

We fit our multiple TE data to the following signal model to estimate the component relaxation times, $$$T_{2}^*$$$, and frequency shifts, $$$\Delta f_k$$$:

$$S(TE)=\sum_{k=1}^3 \rho_{k} \exp(-TE/T_{2,k}^*) \exp(i2\pi \Delta f_k TE), $$

where $$$S(TE)$$$ is the signal at the time of TE, $$$\rho_{k}$$$ represents the signal amplitude of the kth component.

Plots shown in Figure 2 demonstrate that our data at ultrashort TE values in the brain required a 3 component model: a first component with a very short T2* ($$$T_{2,1}^* \approx 0.1$$$ms ) with a negligible frequency shift ($$$\Delta f_1 \approx 0$$$), a second component with a short T2* ($$$T_{2,2}^* \approx$$$ 1-2 ms) with a frequency shift similar to the primary lipid (methylene) resonance ($$$\Delta f_2 \approx -1100$$$ Hz at 7 T), and a third component with a relatively long T2* ($$$> 10$$$ ms) and a a negligible frequency shift ($$$\Delta f_3 \approx 0$$$).

This model was validated in a post-mortem specimen (Fig. 3), where we also observed a very short-T2 component and a short-T2 component with a frequency shift similar to the primary lipid resonance. Since the brain was excised from the surrounding skull, the frequency shifted component originated from within the brain (and was not an artifact resulting from lipids around the brain).

Figure 4 shows typical maps of the fitting model parameters. In the ultrashort-T2 components, $$$\rho_1$$$ was increased in white matter as compared to gray matter, while $$$\rho_2$$$, $$$T_{2,1}^*$$$, $$$T_{2,2}^*$$$ and $$$\Delta f_2$$$ showed little contrast. $$$T_{2,3}^*$$$ also was decreased in white matter and increased in CSF, which is consistent with conventional T2 contrast.

Table 1 shows a ROI-based analysis of the model fitting, showing primarily differences in the 1st and 3rd component fractions between white and gray matter. The first component comprised a relatively large fraction of our UTE signal (21-23% in white matter). Our acquisition was T1-weighted, which will influence these relative component fractions.

The shortest T2 component we observed ($$$T_{2,1}^* \approx 0.1$$$ms) had a shorter relaxation time than previous in vivo UTE measurements ($$$420 \pm 80 \mu s$$$ [1,2]), although some of this differnce can be attributed to using 7T as opposed to 3T. This first component was elevated in white matter, and is likely associated with tightly bound protons in the myelin membranes, proteins, and macromolecules.

The second ultrashort-T2 component ($$$T_{2,2}^* \approx$$$ 1-2 ms) had most notably a frequency shift similar to the methylene lipid resonance, suggesting it is associated with the myelin phospholipid membranes. This result is consistent with previous ex vivo studies, which suggested that methylene protons contributed significantly to ultrashort-T2 components in myelin [3] and also observed a frequency shift in spectroscopic myelin sample experiments [4]. However, to our knowledge, these results represent the first measurement of this frequency shift in vivo. Also, we measured a relatively longer T2* than prior ex vivo work.

The long-T2 component likely includes a lumped measurement of myelin and free water that are observed with conventional MRI techniques. We expect this final component to be a weighted average of myelin, intracellular and extracellular water, and separate characterizations of these components has been performed previously [5,6].

In the brain, we observed an ultrashort T2 component ($$$T_{2,1}^* \approx 0.1$$$ ms) as well as a short T2 component ($$$T_{2,2}^* \approx 1.5$$$ ms) that had a distinct frequency shift corresponding to the methylene proton chemical shift, which to our knowledge has never been observed in vivo. These results are consistent with ex vivo studies, and we also validated our methods in a post-mortem brain specimen to confirm the presence of these two components, which may provide valuable new biomarkers of myelin density, structure, and integrity.