The thalamus relays information to the cortex regions and controls sensori-motor^[1]. Among the sub-structures of the thalamus, nucleus ventro-intermedius(V.im) and ventral posterior lateral nucleus(VPL) are important target regions for DBS^[2]. Accurate identification of the thalamus sub-structures is therefore important for localization of specific nuclei and for the understanding of brain function. However, sub-structures of the thalamus are so small and have intermediate signal characteristics between grey/white matter (GM/WM) it has been difficult to identify them with conventional MR imaging technique. Previous studies have suggested segmenting nuclei by means of the local relaxation times^[3]. Tissue signal suppression using inversion recovery (IR) technique provides a useful tissue contrast particularly tissue is surrounded by the other tissues which have different T₁. However, it has been limited by the difficulty of obtaining high-resolution and high-contrast images at lower filed-strength. Stronger magnetic field strength leads to a higher signal-to-noise ratio (SNR) and allowing for high-resolution MR imaging. Because of increased T₁ at 7T, it is necessary to understand T₁ of thalamus. Several studies measured T₁ at 7T but show large variations^[4-6]. Moreover, measuring T₁ at 7T is challenging due to the long acquisition time. In this work, 1) in-vivo T₁ at 7T are measured using fast T₁ mapping technique, IR prepared EPI sequence with slice reordering^[7,8], in less than 2 min and 2) optimized inversion time (TI) to generate optimal contrast from the thalamus. The optimized 7T images reveal substantially improved anatomical detail of the thalamus sub-structures.Data were collected from 4 controls in a 7T MRI (Siemens; IRB approved) with a 32-receive channel head coil (Nova Medical). Similar data acquisition scheme, an IR-EPI sequence with slice reordering, was used for the T₁ mapping^[7,8]. A non-slice selective adiabatic inversion pulse is applied followed by 5 interleaved slices (Fig. 1A). Due to low SNR of navigator echoes near the null point in IR sequence, data fitting for N/2 ghosting correction has not performed well (Fig. 1B). To improve SNR of navigator echo, it was separately acquired before the start of the IR-sequence. The scan parameters were 80 slices, resolution=1.2×1.2×1.5mm³. To generate T₁ maps, data were processed using a voxel-wised single exponential fitting^[8]. After generating T1 maps, mean T₁ value was measured in GM, WM masks (generated from the T₁ map segmented by SPM5) and the thalamus (manually drawn). TIs for MPRAGE^[9] were estimated using simulation based on measured T₁ values. Then, TIs were jittered to find maximum signal suppression in GM, WM and clearly delineate thalamus sub-structures. Scan parameters were TR/TE = 6000/2.99 ms and resolution = (0.75mm)³.

Compared to the results using conventional navigator-echo acquisition (Fig. 1B), proposed method reveals qualitatively reduced N/2 ghosting artifact (Fig. 1C). A typical T₁ map from a control subject is shown in Fig. 1D. As shown, GM and WM are clearly delineated in T₁ map. Measured mean T₁ values (and standard deviation) in the GM/WM/thalamus from 4 controls were 1851.2 (±153.6), 1178.6 (±336.5) and 1455.5 (±150.259) ms. Mean T₁ value of thalamus is similar with median value of GM and WM (=1514.5ms). Estimated TIs and transverse magnetization plots based on measured T₁ values are shown in Fig. 2A. Estimated TIs for GM-/WM-suppression and thalamus visualization were 1250/800/1000 ms respectively. Figures 2B-D show practically optimized TIs and resulting images. The TI (=930ms), designed to suppress mean T₁ of the thalamus, also suppressed GM/WM border line (Fig. 2C).

Thalamus images using different TIs are shown in Fig. 3. Besides the obvious landmark structures such as the corpus callosum and internal capsule, sub-structures of thalamus were also clearly identified (labeled).

In this study, T₁ values of the in-vivo brain were investigated using IR-EPI sequence to establish parameters for optimal thalamus sub-structures contrast at 7T. The IR-EPI sequence can cover a whole brain volume in less than 2 min. Measured mean T₁ value of thalamus (=1455.5ms) is similar to the median value of WM (=1178.6ms) and GM (=1851.2ms). It can be explained by tissue characteristics of the thalamus, gray matter but parts of thalamus have high axonal density. The optimized results show markedly improved anatomical detail of the sub-structures of the thalamus, including their detailed locations. These evidences suggest it may help us to better understand the brain anatomy in non-invasive way and provide clinically useful information. The estimated TIs show good similarity with practically acquired values but have a difference. It can be explained by individual variation of T₁. The future extension, fast T₁-mapping (using IR-EPI sequence) with high-performance processing computer could provide improved tissue contrast optimization using “real-time” individual T₁-map.