Synopsis

Keywords: Epilepsy, Precision & Accuracy, Stereotactic

Stereo electroencephalography (sEEG) is a valuable tool for localising seizure-onset zones in focal epilepsy. Pre-surgical planning typically combines CT images, which have little spatial distortion, with MRI images, which have improved cerebral tissue contrast. However, the use of CT results in a potentially unnecessary radiation dose to the patient and additional demand on radiology services. In this work, we assess the additional distortion caused by the stereotactic Leksell G-frame in a single healthy volunteer at both 1.5T and 3T. We show that distortion is <1mm in 99.6% of the brain volume at 3T.

Introduction

In stereo electroencephalography (sEEG), depth electrodes are implanted into the brain¹. CT is commonly used to image the head once the stereotaxic frame is in place to minimise distortion. However, using MRI-only approach would remove the exposure to ionising radiation, simplify workflows, and reduce the number of scan sessions required.

Adoption of robotic-assisted electrode placement in sEEG has resulted in increased accuracy compared to optical frameless navigated placement and a reduced margin of safety around blood vessels². This improved accuracy is squandered if the imaging suffers from excessive distortion caused by susceptibility/B0 and eddy-current effects from the frame.

Whilst it is expected that geometric inaccuracies would be greater at higher fields, there are situations where it is more practical to use a 3T scanner or lower field scanners are not available. However, there is still no consensus on the suitability of 3T imaging. Theocharis et al³, evaluating MRI-related distortion for radiotherapy planning, suggested that distortion at 3T may be non-negligible but acceptable <5cm from the MRI isocentre. In contrast, other studies have found the level of distortion due to the presence of the G-frame to be unacceptably high at 3T^4,5.

Here, we assess the distortion caused by the titanium Leksell G-frame. We scan a healthy volunteer with the G-frame and MR indicator box in place and compare the images to those acquired without the frame to assess the additional distortion the frame causes. We repeat the experiment at 1.5T and 3T.

Methods

Image Acquisition
Data were acquired on both a Siemens 3T Vida and Siemens 1.5T Avanto using a Tx/Rx head coil. 3D-T1-weighted MPRAGE acquisitions were performed with the following parameters:

3T: TR/TI/TE=2000/909/3.41ms, isotropic 1mm resolution, 256 partitions, 8° flip-angle, no acceleration, acquisition time 5:31, bandwidth 199Hz/Px.
1.5T: as above except TR/TI/TE=2400/1000/3.71ms, 224 partitions, acquisition time 8:06, bandwidth 180Hz/Px

Following informed consent one participant was scanned with a G-frame, indicator box and G-frame holder (Figure 1 and 2). In the posterior part of the G-frame, titanium pins were inserted upside down, so that the flat back of the pins touched close to back of the skull, separated by laminated card. The front pins were positioned with pointed tips resting just above the participant’s brow, separated by 2 strips of paper tape. The acquisition was repeated without the G-frame, frame holder or indicator box to obtain “reference” images, which should have minimal distortion after the application of the manufacturer distortion correction.

Processing
Images were first bias corrected using SPM12 (https://www.fil.ion.ucl.ac.uk/spm/software/spm12/).

Images acquired with the G-frame and indicator box were registered to their corresponding reference images using a rigid transform followed by a nonlinear registration in niftyreg (http://cmictig.cs.ucl.ac.uk/wiki/index.php/NiftyReg). Displacement maps were generated from the nonlinear component of the registration and the magnitude of the displacements calculated. Following brain extraction (FSL Brain Extraction Tool; https://fsl.fmrib.ox.ac.uk/fsl) histograms of the displacements within the brain were assessed (maximum, peak, median, mean, and standard deviation).

Results

The greatest displacements were observed next to the frame and pins (Figure 3). Maximum displacement magnitudes within the brain (Figure 4) were greater at 3T than 1.5T (1.68 vs 1.20mm). The mean and median distortion were lower at 3T (0.40 ± 0.17mm vs 0.46 ± 0.16mm and 0.45mm vs 0.39mm respectively). The histogram peak occurred at displacements of 0.38mm (3T) and 0.45mm (1.5T).

Field Strength	Histogram Peak (mm)	Median Distortion (mm)	Max Distortion (mm)	Mean distortion ± SD (mm)
1.5T	0.45	0.45	1.20	0.46 ± 0.16
3T	0.38	0.39	1.68	0.40 ± 0.17

Table 1 – displacement observed within the brain volume at 1.5T and 3T


The magnitude of distortion was smaller within the brain, which was far from the frame. The percentage of brain volume with distortion >1mm is 0.36% at 3T and 0.19% at 1.5T. The centre of the thalami (a common target in stereotactic neurosurgery) are approximately 75mm away from the top of the G-frame. In these regions distortion is <1mm at both 3T and 1.5T.

Discussion and conclusion

Images acquired at 3T had a smaller median displacement within the brain volume than at 1.5T (0.39mm compared to 0.45mm at 1.5T). However, there was a greater maximum displacement within the brain volume at 3T than at 1.5T (1.68mm at 3T vs 1.20mm at 1.5T).

As expected, the greatest distortions were near the G-frame and titanium pins. Therefore, distortion may be minimised by positioning the head as superior as possible in the G-frame (within the confines of the indicator box). When targeting structures such as thalamic nuclei, where high accuracy is required, positioning the area of interest central within the G-frame may further help to minimise distortion.

In this work we only assessed distortion with titanium pins, which may cause more severe distortion than other available pins. The use of aluminium or titanium-tipped aluminium pins may further reduce distortion⁴.

In conclusion, we have shown that the distortion caused by the titanium Leksell frame, while non-negligible, is <1mm in 99.6% of voxels within the brain at 3T, using our specific 3D-T1-weighted sequence. Further work is required to assess the cumulative effect of distortion from other sources as well as the effect of distortion on the MR indicator box, which is used for registration in surgery.

Acknowledgements

No acknowledgement found.