White-Matter-Nulled MP-RAGE Predicts Clinical Outcome of Focused Ultrasound Thalamic Ablation for Essential Tremor Jason Su1, Christian Federau2, Thomas Tourdias3, Manojkumar Saranathan4, Casey Halpern5, Jaimie Henderson5, Veronica Santini6, Kim Butts-Pauly2, Pejman Ghanouni2, and Brian Rutt2 1Electrical Engineering, Stanford University, Stanford, CA, United States 2Radiology, Stanford University, Stanford, CA, United States 3Neuroradiology, Bordeaux University Hospital, Bordeaux, France 4Radiology, University of Arizona, Tucson, AZ, United States 5Neurosurgery, Stanford University, Stanford, CA, United States 6Neurology, Stanford University, Stanford, CA, United States I receive support from GE Healthcare. In this project we retrospectively analyze MR-guided focused ultrasound treatment of the Vim in 10 essential tremor patients.

Background: Anatomy Thalamus processes and relays information Nuclei associated sensory signals, motor function, and alertness Essential tremor is treated by targeting Vim Coronal slice of thalamic nuclei from Morel atlas. The equivalent of Vim is highlighted. The thalamus is a central hub for the brain that processes and relays sensory and motor signals between the cortex and subcortex. It contains several subregions or nuclei that are associated with different brain functions. Essential tremor is a movement disorder that afflicts the hands, head, and voice. It can be treated with deep brain stimulation or ablation of the ventral intermediate nucleus or VIM in the thalamus. Morel et al. J Comp Neurol. 1997 Nov 3;387(4):588-630. Schaltenbrand and Wahren Atlas for stereotaxy of the human brain. 1977.

Background: Novel Contrast A: white matter nulled B: gray matter nulled C: CSF nulled (typical MPRAGE) White-matter-nulled MP-RAGE at 7T reveals thalamic structure Tourdias et al. Neuroimage. 2013 Sep 7;84C:534-545.

Background: In-Vivo Delineation 15 in-vivo thalamic nuclei based on Morel atlas 0.69 ± 0.38mm, intra-rater reliability for center of mass Members of our group recently published a work in which they developed a pulse sequence at 7 Tesla called white matter nulled MPRAGE in which the white matter signal is effectively removed from the image. This provides excellent contrast in the thalamus, revealing its nuclei in detail. They were able to manually delineate 15 nuclei based on the Morel atlas for the first time in-vivo, demonstrating a high intra-rater reliability over 6 subjects. Tourdias et al. Neuroimage. 2014 Jan 1; 84: 534–545.

A Retrospective Study of MRgFUS Treatment of ET Towards patient-specific FUS targeting Manually segment and visualize thalamic nuclei Precisely localize FUS ablation of Vim in 14 patients Improve patient tracking and targeting strategy Impact: Direct visualization of nuclei allows precise assessment of ablation targeting This work is a retrospective study of MR guided focused ultrasound treatment of essential tremor in 7 subjects. We use pre-treatment white-matter nulled imaging to first manually segment thalamic nuclei. Then we overlay these on top of post-treatment images to precisely visualize where the ablation occured within the thalamic anatomy. In this way, we aim to improve targeting and better deliver on the promise of truly patient-specific, image-guided focused ultrasound.

Study Overview: 14 patients 7T Scanning 2-4 weeks before treatment WMnMPRAGE 1mm isotropic Not for planning 3T InSightec MRgFUS Locate AC-PC Vim targeted with Talairach atlas Adjustment 3T Scanning Immediately after treatment WMnMPRAGE T2-CUBE The primary novelty here is that we acquire images at 7T and 3T, pre and post-treament, with an imaging sequence that is specifically-tailored to visualize the nuclei of the thalamus. Saranathan et al. Magn Reson Med. 2014 May 29.

Methods: Thalamic Segmentation Manually delineate thalamic nuclei on 7T pre-treatment images Bisect VLP along S-I to obtain Vim In post-processing, we manually segment the thalamic nuclei using the 7T images. The ventral lateral posterior nucleus is typically segmented as a larger body as the boundary between its dorsal half and VIM is invisible. Therefore we simply split it in half along the S-I direction as an approximation. This seems reasonable compared to what is seen in the Morel atlas. Morel et al. J Comp Neurol. 1997 Nov 3;387(4):588-630. Schaltenbrand and Wahren Atlas for stereotaxy of the human brain. 1977.

Methods: Ablation Segmentation Ablation penumbra (manual) Ablation core (automatic: 25% brighter than 99th percentile in rest of thalamus)

Methods: Intrasubject Registration 7T pre-treatment 3T immediate post-treatment We then register the 7T images and the delineated nuclei to the post-treatment 3T images using FSL.

Methods: Intrasubject Registration 3T during treatment 3T immediate post-treatment 1 Targeting adjustment points 2 3 × 4 Final ablation target We also transferred the coordinates of the targeting adjustments to visualize how the search proceeded, shown as colored circles, and where it finally landed for the ablation, the red X.

Study-Specific Template 14 subjects averaged with both original and L-R flipped images Results in a symmetric template Analysis in a normalized brain space has many advantages: Allow allows direct comparison of location and extent between subjects Allows measure of normalized quantities: distance and volume

Methods: Intersubject Registration Intersubject registration is significantly improved by building a study-specific template in ANTS (buildtemplateparallel.sh). Avants et al. Front Neuroinform. 2014 Apr 28;8:44.

Patient F: 7T Pre- Treatment First, here are orthogonal slices of the 7T white matter nulled images.

Patient F: 3T Post- Treatment Then the 3T white matter nulled images immediately after treatment. The ablation zone appears as a salient hyperintensity. Actually, this image contrast has become the favorite of our surgeons for inspecting the ablation.

Patient F: 3T Post with Overlays × is the final ablation target - - is the AC-PC plane We then overlay the thalamic anatomy from 7T and the adjusment points from during the treatment. The red X is the final ablation target and the red dotted line is the AC-PC plane on which initial targeting typically began. We see that the adjustment remained on the AC-PC plane and found the center of the inferior edge of VIM. The ablation as well as its associated edema contains a large potion of the VIM.

Patient E: Best Responder × is the final ablation target - - is the AC-PC plane In this subject, we have a baseline scan after the operation, which shows a more localized ablation than the previous two.

Clinical Correlation at 1-month

Clinical Correlation at 1-month

Clinical Correlation at 1-month

Clinical Correlation at 1-month

Clinical Correlation at 1-month: Best Responder Target We propose new hypothetical targets and analyze these retrospectively Target the ablation zone of our best responder (Patient E) First, we consider the focal zone of our best responding patient and attempt to transfer this region to all the other subjects via nonlinear registration. We evaluated how much of this target was covered by the ablations in each of the subjects and show that this is significantly correlated with the outcome of the subject after 1 month.

Clinical Correlation at 1-month: Intersection Target Target the intersection of all the patients’ ablations Secondly, we construct a probability map by averaging the masks of the FUS affected regions in all subjects using a study-specific nonlinear template. 0% Probability map of FUS-affected tissue across 14 subjects 100%

Clinical Correlation at 1-month: Intersection Target Target the intersection of all the patients’ ablations We threshold this map at 100%, which corresponds to finding the part of the brain that was consistently affected by FUS in everyone. 0% Probability map of FUS-affected tissue across 14 subjects 100%

Clinical Correlation at 1-month: Intersection Target Target the intersection of all the patients’ ablations This makes sense functionally because all patients also consistently had a reduction in hand tremor symptoms at the time of the post scan, immediately post treatment. Coverage of this target is also significantly correlated with the outcome.

These independent hypotheses point to the same target region. Patient E’s Focal Region in white outline Eroded Intersection in solid red Interestingly, these independent hypotheses correspond to the same region! This is initial evidence that future MRgFUS targeting can perhaps be made more accurate and patient-specific with nonlinear registration of WMnMPRAGE targets like this, instead of an affine transformation from a Talairach atlas-based target.

Conclusion and Future Work Coverage of the proposed thalamic target is significantly correlated with outcome WMnMPRAGE permits detailed registration of the thalamic anatomy Offers a patient-specific alternative compared to traditional affine transformation from a Talairach atlas-based target Prospective planning using nonlinear registration

Size vs # Sonications

Can we make a precise ablation core?

Thanks for your interest! Slides available at http://mr.jason.su/