*7.2. T2*

T2w images visualize grey matter as intermediate intensity and white matter as hypointense, although deep grey matter structures can appear even darker depending on the ferromagnetism of their tissue composition. As mentioned, visualization of STN is traditionally achieved with T2w sequences [127–129]. T2w MRI represents transverse relaxation, referring to the amount of time it takes excited protons to lose phase coherence. This dephasing is a tissue-specific process and takes longer for areas with high paramagnetic metal deposition such as iron. As the STN is iron-rich, the contrast is increased, and the nucleus appears hypointense compared to white matter tracts and surrounding grey matter structures. Typically, T2w contrasts within the clinic will come from fast-spin echo sequences that have both a long TE and TR, which are relatively immune to magnetic susceptibility artifacts. However, there is no general consensus as to the optimal sequence required for prime STN imaging. Theoretically, various sequences can achieve the same weighting but vary significantly in terms of their ability to accurately visualize the STN [130]. Moreover, the type of sequence will depend on the field strength, and contrasts are not always analogous across, for instance, 3 and 7 T [131]. Similarly, di fferent MRI vendors will supply similar contrasts via sequences and sequence parameters with di fferent names, making it di fficult to draw comparisons between them [50,132,133].

### *7.3. T2\* and Susceptibility-Based Contrasts*

Traditional clinical T2w sequences su ffer from low signal and contrast. An alternative contrast that can be used to image the STN directly comes from 3D gradient echo (GRE) sequences, which can be used to create T2\* images. Typically, GRE sequences will include a low FA, long TEs, and long TRs. Moreover, gradients are applied to initiate dephasing, as opposed to an RF pulse in traditional spin echo sequences [109,134]. These gradients do not refocus field inhomogeneities such as RF pulses do. Therefore the T2\* contrast arising from GRE reflects magnetic field inhomogeneities caused by the dephasing of neighboring areas that occurs at di fferent rates, and further interact with the signal of adjacent voxels [135]. As GRE sequences assess macroscopic intervoxel and microscopic intravoxel magnetic susceptibilities, it is important to adapt sequence parameters according to the tissue of interest [136]. The tissue characteristics of the STN undergo PD-specific changes, such as dopaminergic denervation and excessive iron deposit, which require adjusted parameters such as TE for optimal contrast [137,138]. Similarly, iron increases with normal aging requires di fferent adaptations to TEs [139]. GRE sequences also incorporate multiple echoes to account for di fferences in magnetic susceptibility across tissues. Further, susceptibility e ffects are stronger for smaller voxel sizes as the dephasing is reduced [135]. This makes T2\* imaging more appropriate for higher field strength MR, as smaller voxel sizes can be achieved with faster acquisition times [130,140]. These T2\* images can be further processed to create quantitative maps that will be discussed in later sections.

Alternatively, susceptibility weighted images (SWI) can be created from T2\*-based sequences by independently processing magnitude and phase images. Magnitude images reflect the overall MR signal, and their corresponding phase image contains information about field inhomogeneity, di fferences in local precession frequencies, and motion [141]. Phase images were largely discarded before the implementation of SWI as they require complex unwrapping, referring to the extraction of their original numerical range, which is constrained in the outputted image to [−π, <sup>+</sup>π] [142]. However, phase can be used to visualize information that would otherwise be barely visible in magnitude images. Small structures result in field variations with high spatial frequencies, which can be used to enhance contrast by applying a high pass filter. The resulting SWI image is the product of multiplying the phase mask with the magnitude image [142–144]. It remains somewhat controversial to what extent SWI signal increases from 1.5 T to 3 T MRI. Moreover there is little evidence for increased accuracy for SWI at 3 T compared to classic T2 imaging [145]. However, SWI is significantly more accurate compared to traditional contrasts at higher field strengths [146–148]. GRE-based sequences and T2\* contrasts can provide more detail regarding the shape, surface, and location of the STN compared to standard T2w spin echo-based sequences. This could translate to more accurate DBS targeting if it were used for preoperative planning. Improvements can refer to a smaller deviation between planned and actual lead location, a reduction in reimplantation or removal requirements, increased clinical efficacy, or decrease in associated side e ffects. However, the use of T2\* contrasts and UHF-MRI remains widely debated and requires further validation [37,70,144,148–150].

We attempted to use a T2\*-based UHF-MRI with a GRE-ASPIRE sequence [151] on a 7 T Siemens MAGNETOM system (Siemens Healthcare, Erlangen, Germany) for STN DBS planning in PD patients. The 7 T T2\* scan consisted of a partial volume covering the subcortex, obtained with multiple echoes (TE1–4 = 2.47, 6.75, 13.50, 20.75) and 0.5 mm isotropic voxel sizes in just under 8 min. This was overlaid with a 3 T T2w turbo field echo sequence obtained on a 3 Tesla Phillips Ingenia system, with a single TE of 80 ms and voxel sizes of 0.45 × 0.45 × 2 mm, and an acquisition time of around 6 min. When merging the 3 T and 7 T data, the STN appeared elongated along the posterior direction on 7 T. The optimal target coordinate appeared more superior, posterior, and lateral on the 7 T image than the optimal coordinate on 3 T. Here, the 7 T coordinate was used as the posterior test site sampled with MER was used as a target for DBS surgery. Typical STN activity was not observed, although intraoperative behavioral testing revealed that patients would exhibit a beneficial clinical e ffect. Such a finding may be explained by the fact that the test electrode was instead stimulating white matter fibers exciting the STN, such as the fasciculus lenticularis or medial fiber bundles. It is, however, unclear as to

whether this discrepancy in optimal STN coordinate is due to errors in registration across field strength, smoothing factors and interpolation automatically applied by the pre-operative planning system that reduced the resolution of the 7 T data, magnetic field inhomogeneity, or geometric distortions of the T2\* image. The issues regarding image correction and manipulation are discussed in later sections. It is entirely plausible that the discrepancy in optimal target location across field strength was due to human error, and the operating surgeons perhaps were not used to interpreting the high-resolution susceptibility-based images. Therefore, factors other than contrast and sequence type can influence the usability and accuracy of susceptibility-based imaging for neurosurgical applications.

It is important to note that the sequences described in this specific instance are not standardized across centers, and scanner vendors, field strengths, contrasts, and sequence parameters, even within the same sequence type, will di ffer across DBS centers and research institutes. This makes a direct comparison across the quality and replicability of MRI scans very di fficult, and unless systems are harmonized, interpretations should be site-specific. See [86,130] for a comprehensive review on sequences used for imaging the STN.
