Direct Visualization of Anatomic Subfields within the Superior Aspect of the Human Lateral Thalamus by MRI at 7T

BACKGROUND AND PURPOSE: The morphology of the human thalamus shows high interindividual variability. Therefore, direct visualization of landmarks within the thalamus is essential for an improved definition of electrode positions for deep brain stimulation. The aim of this study was to provide anatomic detail in the thalamus by using inversion recovery TSE imaging at 7T. MATERIALS AND METHODS: The MR imaging protocol was optimized on 1 healthy subject to segment thalamic nuclei from one another. Final images, acquired with 0.52-mm2 in-plane resolution and 3-mm section thickness, were compared with stereotactic brain atlases to assign visualized details to known anatomy. The robustness of the visualization of thalamic nuclei was assessed with 4 healthy subjects at lower image resolution. RESULTS: Thalamic subfields were successfully delineated in the dorsal aspect of the lateral thalamus. T1-weighting was essential. MR images had an appearance very similar to that of myelin-stained sections seen in brain atlases. Visualized intrathalamic structures were, among others, the lamella medialis, the external medullary lamina, the reticulatum thalami, the nucleus centre médian, the boundary between the nuclei dorso-oralis internus and externus, and the boundary between the nuclei dorso-oralis internus and zentrolateralis intermedius internus. CONCLUSIONS: Inversion recovery–prepared TSE imaging at 7T has a high potential to reveal fine anatomic detail in the thalamus, which may be helpful in enhancing the planning of stereotactic neurosurgery in the future.

phology can show significant interindividual variability, 3 direct visualization of intrathalamic anatomy is necessary for a more precise patient-specific planning of brain electrode positions.
Besides the small size of some thalamic nuclei, the main reason for the lack of contrast is the similarity of MR imaging-relevant tissue parameters such as relaxation times and/or proton densities of adjacent thalamic structures. However, as long as differences in contrast-relevant tissue properties exist, their delineation by an appropriate MR imaging technique is mainly a question of SNR. Several years ago, the wide availability of 3T MR imaging scanners triggered a series of studies aiming to visualize internal substructures of the thalamus, 4-11 extending the research done at 1.5T. [12][13][14][15][16] Although noticeable progress was made, the visualization of thalamic nuclei with 3T MR imaging for DBS surgery is not a clinical routine to date. It seemed reasonable to assume that the SNR boost offered by MR imaging at 7T could further improve the visualization of thalamic nuclei. Indeed, the first encouraging results have been reported by using SWI, quantitative susceptibility mapping, and MPRAGE imaging. [17][18][19] This feasibility study aims to investigate the usefulness of inversion recovery turbo-spin-echo (IR-TSE) MR imaging at 7T in revealing anatomic detail within the thalamus. This task comprises both the optimization of the MR imaging acquisition and the assignment of visualized structures to thalamic anatomy.

MATERIALS AND METHODS
We examined 5 healthy subjects (2 men; 21-28 years of age) in accordance with stipulations put forth by the local ethics committee. All subjects gave written informed consent before the experiment.
MR imaging was performed at 7T (Siemens, Erlangen, Germany) with a 32-channel head coil (Nova Medical, Wilmington, Massachusetts). Each subject's head was immobilized by using cushions. Optimization of the IR-TSE protocol was done with a subject who had been specifically trained to keep his head still during prolonged scanning times. Scanning parameters of the final IR-TSE acquisitions with high in-plane resolution were the following: transversal-section orientation parallel to the anterior/ posterior commissure (ACPC) plane; section thickness, 3-mm; in-plane resolution, 0.5 2 mm 2 ; turbo factor, 5-7; TR, 3-3.2 seconds; TE, 10 -14 ms; TI, 480 -500 ms; bandwidth, 140 -203 Hz/ pixel; 1 section or 10 sections with 2 interleaved concatenations; averages, 2-5; scanning time, 13 minutes 45 seconds through 19 minutes 10 seconds. To estimate longitudinal relaxation time, T1, we acquired a series of the IR-TSE protocol, measuring a single section at a section position of 10 mm cranial to the ACPC plane with a TR of 3.2 seconds and a varied TI of 25, 250, 500, 750, 1000, 1250, 1500, and 2000 ms with an in-plane resolution of 0.6 2 mm 2 . The T2 map was measured at the same section position with a series of spin-echo measurements with a constant difference between the TR and TE. 20 For the sake of comparison, TSE images were acquired as well with the following parameters: 2.5-mm section thickness and 0.5-mm intersection gap; 0.5 2 -mm 2 in-plane resolution; turbo factor, 5; TR, 3 seconds; TE, 15 ms; bandwidth, 131 Hz/pixel; 1 concatenation with an inferior saturation band; averages, 3; scanning time, 13 minutes 30 seconds. The other 4 subjects were imaged for the first time in a 7T scanner. To reduce the risk of motion-induced image degradation, we selected an IR-TSE protocol with a shorter scan duration and a larger voxel volume (2-mm section thickness; 0.7 2 -mm 2 in-plane resolution; turbo factor, 7; TR, 3.3 seconds; TE, 10 ms; TI, 500 -700 ms; bandwidth, 186 Hz/pixel; 8 sections with 2 interleaved concatenations; averages, 2; scanning time, 10 minutes 2 seconds. To optimize the SNR of the applied sequences, we adjusted the global amplifier reference to ensure that the nominal flip angles of 90°and 180°applied in the thalamus. The T1 map was fitted voxelwise to the signal-intensity equation: I ϭ c {1-2 exp(ϪTI/T1) ؉ exp[Ϫ(TRϪTE last )/T1]}, where TE last denotes the echo time of the last echo in the echo train and c is a scaling factor. The T2 map was fitted to a single exponential decay. MR images were compared with labeled photographs and drawings in stereotactic brain atlases 3,21,22 to assign visible differentiation to known thalamic anatomy. Apart from a few exceptions, we used the terminology introduced by Hassler. 1 Figure 1 shows axial MR images through the thalamus at a position of about 10 mm dorsal to the ACPC plane, with the corresponding section of the atlas of Schaltenbrand and Wahren 21 at position H.d ϩ7.0. At first glance, a high degree of similarity between the IR-TSE image (TI ϭ 500 ms) and the T1 map (calculated from a series of MR images with different TIs) is noticeable. However, more striking is the richness of visible anatomic detail in the lateral thalamus. For illustrative purposes, some salient landmarks are marked in white (Fig 1, bottom row). Arrowheads point to the mamillothalamic tract, paired arrowheads indicate the stria medullaris, and arrows point to the nucleus centre mé-  To understand why IR-TSE is beneficial compared with proton-density-weighted TSE imaging as a method of differentiating the D.o.i from the Z.im.i, we modeled the contrast behavior

DISCUSSION
The comparison of MR images with photographs and drawings of stereotactic atlases revealed that the Schaltenbrand and Wahren atlas 21 seems to be the most suitable one. Its myelin-stained plates appear similar to the presented MR images with hypointense myelin-rich and more hyperintense myelin-poor tissues. Furthermore, the overlayable transparencies with drawn borders of the different thalamic compartments based on their cytoarchitecture notably facilitate anatomic assignment. The remarkable Morel atlas 3 shows photographs of the variety of applied stainings only exemplarily. For image comparison and assignment, however, we rate camera lucida drawings less suitable than labeled photographs of stained sections. Unfortunately, the atlas of Mai et al 22 presents only coronal views in high magnification with elaborate anatomic labeling but no sagittal and axial ones.
Despite the high degree of differentiation, the proposed assignment of areas with specific MR imaging contrast to anatomically described tissues remains somewhat ambiguous. One possible reason for discrepancies is the markedly different nature of the images under consideration. MR images integrate  tissue properties over the applied section thickness and in-plane voxel resolution. In contrast to the presented MR images with 2-to 3-mm section thickness and 0.5 2 -0.7 2 mm 2 in-plane resolutions, photographs in brain atlases based on much thinner histologic sections show the distribution of different tissue compartments for the specified position with extremely high resolution. Consequently, the MR images are more similar to an average of neighboring sections in anatomic atlases than to the MR imaging equivalent of 1 specific section. This observation complicates anatomic assignment, especially if anatomy changes considerably over short distances in a direction orthogonal to the sections. One exam-ple is the nucleus zentrolateralis intermedius. Because it is a thin layer between the larger ventral and dorsal intermediate nuclei, 1 we cannot exclude the observation that the field here assigned to Z.im.i belongs to one of the latter subfields. Nevertheless, we decided to adopt the labeling of the atlas photograph with the best correspondence to the MR image.
Besides the Schaltenbrand and Wahren atlas, 21 the partition of the lateral thalamus parallel to the La.m and internal capsule appears even more conspicuous in the myelin staining in Fig 46 in the work of Hirai and Jones. 25 Here the nuclei are named ventral lateral anterior (VLa) and ventral lateral posterior (VLp). For an elaborate analysis of which nuclei as defined by Hassler 1 correspond to the nomenclature of Hirai and Jones and a critical view on Hassler's very large number of subdivisions, we refer to the tables and discussions given elsewhere. 23,[25][26][27][28] To successfully delineate the anatomic subfields in the lateral dorsal thalamus, T1-weighting is essential. Ultimately, an inversion recovery approach proved to be most suitable. The popularity of gradient echo imaging at 7T [17][18][19] is driven, to a high degree, by the possibility of covering the whole brain in a couple of minutes in compliance with specific absorption rate regulations. The applied IR-TSE sequence, with adiabatic inversion and full 180°refocussing pulses, allows us to measure only a small brain slab within legal specific absorption rate limits. The resulting 8 -10 possible sections are sufficient to cover the region of interest. To measure the planned number of sections, occasionally, we were forced by the specific absorption rate limit to increase the TR from the nominal 3 seconds up to 3.3 seconds. Whereas the concomitant slight increase in contrast is welcomed, the prolonging of scanning time is disadvantageous. For IR-TSE imaging, the application of specific absorption rate-reducing techniques is possible yet untested for the presently proposed application.
Unlike the specific absorption rate, CSF pulsation is a serious obstacle. Sharp and contrast-rich images can be reliably acquired in dorsal thalamic regions where the rather small flow distortions are controllable by an appropriately chosen phase-encoding direction. However, in a manner strongly dependent on the subject, IR-TSE images around the ACPC plane can be heavily distorted by flow artifacts arising from the strong pulsatile CSF flow in the third ventricle. Often these cannot be suppressed sufficiently by the flow compensation options of the product TSE sequence. Moreover, flow-synchronous triggering is far from simple-if not impossible-for inversion recovery-prepared MR images. Currently, CSF pulsation artifacts are the reason why only images of the dorsal aspect of the thalamus are described here.
In this regard, 3 recent studies explicitly described the delineation of thalamic nuclei at 7T. [17][18][19] Abosch et al 17   The MR images were acquired with a resolution of 0.5 2 ϫ 3 mm 3 (IR-TSE, TI ϭ 500 ms) and 0.5 2 ϫ 2.5 mm 3 (TSE). In the bottom row, which is identical to the top row, arrows point to vessels, and arrowheads, to bright spots. The latter are flow artifacts in the medial thalamus originating from branches of thalamic veins. Note the nearly complete lack of contrast inside the lateral group of thalamic nuclei, the very faint appearance of the external medullary lamina, and the pronounced appearance of vessels in the TSE image. plane. Whereas Abosch et al did not show sections located more dorsally, images presented by Deistung et al showed no indication of parcellation in the dorsal aspect of the lateral thalamus. Also the La.m appears as a rather broad and diffuse band in the quantitative susceptibility map, notwithstanding the extremely high 0.4 3mm 3 spatial resolution of the underlying MR images. The white matter-nulled MPRAGE images of Tourdias et al 19 explicitly indicated the nuclei ventral lateral anterior and posterior in the dorsal thalamus; however, they did not depict a structure similar to the D.o.i/Z.im.i boundary shown here. The coarser appearance of the La.m and of boundaries between the nuclei in their axial images is probably a consequence of the isotropic 1-mm 3 resolution. All these points are not criticism. We just intended to underline the importance of determining the most suitable MR imaging approach to distinguish a particular thalamic nucleus from adjacent tissues.
Target definition is a core element in the complex process necessary for precise stereotactic implantation of brain electrodes. More precise localization of the supposed optimal target for therapeutic interventions cannot balance the impact of other sources of error such as image resolution or the mechanical accuracy of stereotactic systems. However, it reduces the total error of the procedure and thus promotes a more comprehensive understanding of therapeutic outcomes. In that regard, 2 recent studies 29,30 investigated possible issues for neurosurgical targeting due to higher geometric distortions at 7T. Without dissent, they reported relatively small distortions in the central brain regions and concluded that targeting is feasible with 7T imaging. Without question, the assumed advantage of the clear visualization of intrathalamic anatomy for DBS planning has yet to be validated. We expect that among others, the La.m, the boundary between the nuclei D.o.i and Z.im.i, the boundary between the nuclei D.o.i and D.o.e, and the dorsal aspect of the centre médian are valuable landmarks. Because the boundary between the D.o.i and Z.im.i is located only a few millimeters dorsal to the anterior margin of the nucleus ventrointermedius internus, it has the potential to promote a more accurate DBS planning in patients with tremor. The successful transfer of the imaging protocol to patient scanning is strongly bound to an effective suppression of any head movement. A thorough comparison of the recorded images of the subject trained to keep his head very still, on the one hand, with the images of the other 4 subjects, on the other hand, suggests that residual head motion may explain the sometimes less compelling visualization of subfields in the lateral thalamus. Methods of mitigating these artifacts used in routine MR imaging investigations of patients with motor disorders, such as the use of sedative medications or MR imaging-compatible stereotactic head frames, are not, in our view, feasible options in a 7T context. Instead, prospective motion-correction technologies 31 that do not require a tight fixation of the head present a promising alternative. Moreover, prospective correction of microscopic head motion has already been demonstrated to be fully compatible with 7T MR imaging. 32

CONCLUSIONS
This feasibility study demonstrates that IR-TSE-based MR imaging at 7T can reveal a high degree of anatomic detail in the dorsal thalamus. Even if none of the here-visualized subfields are cur-rently a relevant DBS target, the more general hope for the future is that precise visualization of well-defined internal thalamic landmarks will improve the definition of target point coordinates for stereotactically guided implantation of DBS electrodes.