Inflammatory disorders of the central nervous system such as multiple sclerosis often involve the spinal cord (SC) besides the brain. However, imaging of the SC poses more technical challenges due to the smaller and more mobile structure. Additionally, it is often observed that the SC exhibits a quite low gray matter (GM) and white matter (WM) contrast in standard MRI protocols. In order to develop optimized SC MRI sequences in the future and to facilitate a further tissue characterization of the SC, a dedicated and precise T1 mapping protocol for the SC was developed.

Fast T1 relaxometry with high in-plane resolution was facilitated by using ultra-fast 3D RF spoiled gradient echo sequences with Half-Fourier reconstruction (VIBE): FOV = 104mm x 208mm, resolution = 0.4mm x 0.4mm x 5mm, 8 slices in one slab, 100% slice oversampling, GRAPPA parallel acquisition technique with acceleration factor 2, TE = 7.38ms, TR = 13.3ms, no signal averaging per acquisition, TA = 50sec. The center of the transversal slab, located between the cervical vertebrae C5 and C6, was orthogonally orientated to the course of the spinal cord. The VIBE acquisitions were performed alternatingly with flip angles of 5 deg and 15 deg until 42 acquisitions were completed for each flip angle resulting in a total experiment time of 2x42x50sec = 70min.

The two differently T1 weighted series of 42 image datasets each were co-registered using elastic registration algorithms of the Elastix toolbox [2]. Based on the two resulting averaged datasets of different T1 contrast, quantitative T1 maps of the SC were calculated. For this, the signal equation for RF spoiled gradient echo (and therefore VIBE) sequences was analytically solved for T1 according to Ref. [3]. In order to achieve precise T1 values, additional mapping of the B1⁺ RF transmit field was performed as published by Ganter [4]. The resulting B1⁺ data was used for correcting the RF flip angle to the real (effective) value in the human body. All measurements were performed on a 3T whole-body MR system.

Regions of Interest (ROI) were drawn on the calculated T1 maps: For GM at the cornu anterius and cornu posterius, left and right, thereby omitting the commisura grisea with the canalis centralis, and for WM in the central part between the corni posterii (i.e. the sulcus medianus posterior). The two central slices of the slab were evaluated, the peripheral slices of the slab discarded due to the immanent slab excitation profile.

Figure 1 illustrates mean images of the VIBE acquisitions in one of the two evaluated slices for both the 5deg and 15deg acquisitions. Figure 2 demonstrates a resulting T1 map. Both Fig. 1 and Fig. 2 were extremely zoomed into the SC, since it is the only region of interest. Table 1 presents the assessed T1 values for GM and WM of the SC.

The present work quantified T1 relaxation times in GM and WM of the SC. Since elastic registration methods were employed, the impact of such techniques on the resulting T1 values could be of concern. However, initial investigations with different frame parameters showed that the impact on the T1 values was practically negligible compared to the "natural scattering" of the T1 values.

It is worth to note that the mean results determined for WM-SC (976ms) and GM-SC (1154ms) deviate notably from the corresponding T1 values found for the cerebellum at 3T: approx. 900ms (WM-cerebellum) and approx. 1500ms (GM-cerebellum) [5]. As a matter of fact, the measured small T1 difference for the SC of only 178ms may partially explain the relatively low contrast between GM and WM frequently observed in SC imaging.

In the future, the knowledge of these precise T1 values will allow to set up dedicated SC imaging protocols with an optimized GM-WM contrast (e.g. magnetization prepared rapid gradient echo, MPRAGE). These optimized SC imaging protocols may be expected to lead to further benefit in the MRI based investigation and diagnosis of SC pathologies.