The North American Neuromodulation Society (NANS) and the Neural Interfaces Conference (NIC) is holding a joint scientific conference on June 25–29, 2016, in Baltimore, MD.
This conference convenes a diverse group of scientists, engineers, and clinicians representing the basic and applied science aspects of neural interfaces and neuromodulation. The goal of the conference is to foster collaboration between these groups and provide an in-depth overview on the research and development of implantable medical devices and techniques along with their eventual integration into clinical practice.
Butson Lab members recently presented eight posters at the NANS/NIC Conference. A poster from Daria (Nesterovich) Anderson was one of four selected for a "Platform Presentation".
Research presented included new techniques and advances on neuromodulation and deep brain stimulation for the treatment of movement disorders and pain. Please use the navigation tool on the left to find out more about our exciting research.
Are Directional Deep Brain Stimulation Leads Safe? An Updated Assessment of Stimulation Safety
Ashley Kapron1; Daniel Merrill; Christopher Butson, PhD1
1 University of Utah
Directional deep brain stimulation (DBS) leads are now being assessed for clinical usage. These novel lead designs may improve the ability to selectively stimulate target regions and avoid adjacent structures associated with side effects. With less surface area, directional electrodes may experience higher charge density than traditional cylindrical electrodes, particularly at edges where charge density is concentrated. Since elevated charge density has been previously implicated in neural damage, there is concern that the local charge density at neighboring edges may exceed safety standards, especially during bipolar stimulation of adjacent contacts. Therefore, the objective of this study was to compare the local charge densities at the edges of traditional and directional DBS electrodes to safety limits described in literature.
Computational finite element models were created in COMSOL 5.2 to predict and visualize charge densities on cylindrical (Medtronic 3389) and directional (directSTN Acute) contacts (electrode shaft 1e-6 S/m; electrode contacts 1e6 S/m). Electrodes were centered in a cubic volume conductor representing the brain (edge length 0.1 m, 0.2 S/m) and were surrounded by a 0.5 mm thick encapsulation layer (0.07 S/m). Charge density (D) associated with monopolar and bipolar stimulation over a range of charge per phase (Q) settings was calculated over the entire cathodic electrode contact as well as a 0.2 mm wide section at the edge closest to the adjacent anodic contact. Monopolar and bipolar results were compared within/between electrodes and to safety limits proposed by Shannon (IEEE Trans Biomed Eng, 1992): log(D) = k – log(Q).
Results The overall charge density was 5.0 times greater in the directional contact compared to the cylindrical contact, but only 3.8 times greater at the 0.2 mm edge section during bipolar stimulation. While charge density over the entire contact did not differ between monopolar and bipolar settings, bipolar stimulation increased the charge density at the 0.2 mm edge section by 10% in the directional contact and 17% in the cylindrical contact relative to monopolar stimulation (Figure 1).
Bipolar stimulation results in increased charge density at the edges of adjacent contacts. More importantly, the decreased surface area of directional leads causes a substantial increase in both overall and edge charge density compared to traditional cylindrical leads. Therefore, a limited operating range of stimulation settings may be required to prevent neural damage when using directional leads. These computational results motivate the need for in-vivo studies confirming the safety of novel directional lead designs.
Nonlinear Atlas Creation for Retrospective DBS Cohort Analysis
Gordon Duffley1; Kelly Foote, MD2; Michael Okun, MD2; Christopher Butson, PhD1
1 University of Utah 2 University of Florida
The locations of stimulating electrodes during all forms of neuromodulation therapy have been correlated with clinical benefit. However, when performing retrospective cohort analysis, many studies use simple registration techniques, often an automated affine registration to MNI or Talairach coordinate space. The field of image registration has developed techniques for creating synthetic atlases from a cohort of patients, which have been shown to create coordinate spaces where spatial comparisons are more accurate. In this study, we analyze how spatial location of deep brain stimulation leads changed when comparing traditional registration methods to a nonlinear registration to a synthetic atlas.
In order to compare methods of mapping data into a common coordinate space, we used imaging from 22 unilateral DBS patients (14 STN targeted, 8 GPi targeted) from the University of Florida Center for Movement Disorders. All had Medtronic 3387 leads implanted. We identified lead locations in native patient space from postoperative CT and represented the patient’s lead as a set of 5 points (tip of the electrode and center of each contact). We mapped points representing the DBS leads to MNI305 space by applying the transform generated by FreeSurfer after running each preoperative MRI through the pipeline. We created our synthetic atlas from T1 imaging using ANTs antsMultivariateTemplateConstruction algorithm. To compare the coordinate systems, we used the python packages scikit-learn, numpy, and scipy.
The ranges of the points representing our electrodes for STN patients in MNI305 were enclosed in an ellipsoid of with principle axes of 9.53mm, 12.93mm, and 15.31 mm, and 6.92mm, 13.27mm and 16.91mm for GPi patients. The equivalent ellipsoids in the non-linear atlas had principle axes of 8.72mm, 7.05mm, and 15.97mm for STN targeted leads, and 6.19mm, 5.29mm and 13.56mm for STN targeted patients. To further highlight that the different coordinate spaces resulted in drastically different relative lead locations, when running a k-means clustering on the points that represented our electrodes, in our nonlinear atlas space, the two clusters split the points between GPi and STN targeted leads, where MNI305 space did not (Figure 1).
Relative lead locations are drastically different in our synthetic atlas space compared to FreeSurfer’s registration to MNI305 space. Given the mathematical basis of nonlinear atlas approaches and the clean clustering in our synthetic atlas space, it supports the idea that nonlinear atlas approaches should be used for future cohort analysis.
Optimized Programming Algorithm for Cylindrical and Directionally Segmented Deep Brain Stimulation Electrodes
Daria Nesterovich; Braxton Osting, PhD; Alan Dorval, PhD; Christopher Butson, PhD
University of Utah
Deep brain stimulation (DBS) programming is a complex process likely to become more complex with the introduction of leads with larger numbers of contacts. We developed an automated programming algorithm to optimize DBS parameter selection for targeted neural activation in a patient-specific manner. The purpose of this study is to assess algorithm performance by applying it to conventional and directional electrode geometries.
We used finite element models to solve the bioelectric field problem for a conventional 4-contact DBS lead (Medtronic 3387) and three directional leads: the directSTNacute (Pollo et al., 2014), the segmented lead, (Buhlmann et al., 2011), and the Sapiens electrode (Decré et al., 2013). As shown in Figure 1A, a 3D rendering of the stimulation target area and fiber tractography were generated from patient MRI and diffusion tensor imaging (DTI). We used values derived from the Hessian matrix of voltage second derivatives as an estimate of neural activation to maximize stimulation in the STN and limit activation of axon fibers in the internal capsule.
We have demonstrated an ability to program the electrode to stimulate a target area while avoiding neural tracts that may be responsible for side effects. We tested that the algorithm settings were robust by adjusting the magnitude of optimized electrode settings as shown in Figure 1B. Optimal parameter settings from the algorithm show STN activation while limiting stimulation outside the target area and the internal capsule.
We have developed a method of optimization that can be applied to the clinical electrode and additional complex DBS lead designs. Deviation away from the optimized parameters showed more stimulation outside the target area or activation of neighboring fiber tracts. A real-time, patient-specific, automated programming algorithm may increase efficiency and positive outcomes of clinical DBS programming, as well as enable the use of more complex lead designs, which are likely to be too complex for manual programming.
Deep Brain Stimulation for Neuropathic Pain: Connectivity Analysis within the Sensory Thalamus
Yagna Pathak1; Guillermo Horga, MD PhD1; Gordon Duffley2; Nora Vanegas-Arroyave, MD1; Christopher Butson, PhD2; Sameer Sheth, MD PhD1
1 Columbia University Medical Center 2 University of Utah
Neuropathic pain affects nearly 3% of the global population. Neuromodulatory interventions such as deep brain stimulation (DBS) target the periaqueductal/periventricular gray (PAG/PVG) or ventral posterior medial/lateral (VPM/VPL) thalamic nuclei. However, the mechanism of action involved in pain control remains unclear. In this study, we aim to characterize the network properties that correspond with effective clinical response resulting from VPM stimulation.
A 34-year-old male underwent right-sided DBS in the VPM and PVG for refractory left maxillary-division chronic neuropathic facial pain. We acquired pre-op structural (3T) MRI and diffusion weighted imaging (DWI) (64-directions, B-value=1000s/mm2). We determined contact locations by co-registering the pre-op MRI with a post-op CT. Volume-of-tissue activated (VTA) was defined for each contact at the two targets. Finite element methods were applied to model the VTAs while accounting for stimulation parameters and diffusion-tensor based conductivities of the surrounding tissue. Deterministic tractography seeded at each VTA was conducted in FSL and DSI studio to contrast effective versus ineffective stimulation settings.
Low-frequency stimulation (30Hz) was tested at both targets. The patient demonstrated better clinical response with VPM compared to PVG stimulation. Among the 4 VPM contacts (c0-c3), stimulation at c1 and c2 corresponded with better analgesic response. Preliminary results from tractography indicated increased fiber density and FA at the dorsal VTAs compared to the ventral ones—c0:17 tracts, FA=0.37±0.12; c1:6 tracts, FA=0.44±0.16; c2:48 tracts, FA=0.43±0.15; c3:90 tracts, FA=0.43±0.16. Most VTAs (c0, c2, c3) projected ventrally to the right medial lemniscus, right superior and inferior cerebellar peduncles. Projections from c1 and c2 also passed through secondary sensory cortex (SII:BA 4 and 5).
Altered neuronal activity in the thalamic nuclei may be involved in the mechanism of pain. Specifically, neuroimaging and lesion studies have supported the role of VPM in pain processing. It has been suggested that suppression of neuronal hyperactivity projected from the trigeminal system to the thalamus contributes to the alleviation of pain. Our results support this notion as fibers passing through most VTAs included the medial lemniscus, which contains projections from the trigeminal nerve and relays sensory information to the VPM. Further, projections from c1 and c2 pass through SII, an area that has a discriminative role in pain processing. These results also suggest that anatomical connectivity is a crucial piece in understanding the network that is involved in pain processing. Further, this network can be leveraged to identify nuanced targets for future DBS-pain studies.
Interactive Meshing and Simulation of Deep Brain Stimulation With Patient Specific Models
Andrew Janson1; Jonathan Baker2; Nicholas Schiff2; Keith Purpura2; Christopher Butson, PhD3
1 Scientific Computing and Imaging Institute 2 Weill Cornell Medical College 3 University of Utah
Deep brain stimulation (DBS) is an effective therapy for movement disorders such as Essential Tremor and Parkinson’s disease and is a potential therapy for several other disorders. DBS is becoming more complex: novel lead designs are being introduced with directional electrode contacts, and in some cases multiple leads are being implanted in close proximity in a single subject. Finite element bioelectric field models have been used successfully to predict the effects of DBS, but until now these have been time-consuming and computationally costly to generate. We have developed Virtual DBS: an interactive system to simulate and visualize the effects of DBS in patient specific models for arbitrary electrode geometries and configurations of anodes and cathodes.
Virtual DBS uses an image processing pipeline has been developed to incorporate structural MRI and diffusion weighted imaging (DWI) into an interactive, patient specific model of DBS. The DWI data is used to estimate brain tissue conductivity. Multiple DBS leads can be interactively placed inside of the patient model; boundary conditions can be applied to simulate anodes and cathodes. Changing the lead position or trajectory triggers the generation of a new finite element mesh and solution of the bioelectric field problem.
The Virtual DBS system can produce a finite element mesh, solve the bioelectric field problem, postprocess data, and visualize the results in under 10 seconds. Users are able to move the DBS lead within the patient model and can rapidly explore different stimulation parameters to predict the effects of DBS on surrounding nuclei and fiber tracts. This tool also allows for the placement of multiple leads within a single hemisphere to allow for current steering by varying the anodes and cathodes on each lead. For example, Figure 1 shows current steering with three leads and the activating function on target fiber pathway calculated from the tripolar electric field.
We anticipate that this simulation tool will enable interactive exploration of DBS effects within and among subjects. Further, we anticipate that a combination of modeling and postoperative clinical analysis will elucidate the mechanisms of DBS in ways that were not previously possible.
Rapidly Reversible Behavioral Arrest During Fasciculus Retroflexus Deep Brain Stimulation in a Healthy Nonhuman Primate
Jonathan Baker1; Jae-Wook Ryou, PhD; Christopher Butson, PhD; Nicholas Schiff; Keith Purpura, PhD
1 Weill Cornell Medicine
Sometimes the best response is none at all. Rapid cessation of purposeful movement is frequently observed when a stimulus-response pair is associated with an aversive outcome, for example a lower than expected reward, or a painful or threatening stimulus. Behavioral arrest of ongoing movement is a general strategy for survival in all vertebrates and is proposed as one of the primary functions of the lateral habenula, an evolutionarily conserved epithalamic structure that integrates broad cortical and subcortical inputs to modulate monaminergic midbrain structures that regulate motivation and motor outflow. The primary descending output of the habenula, the fasciculus retroflexus, is robust fiber bundle that traverses the parafascicular nucleus, part of the centromedian-parafascicular (CM-Pf) complex and the caudal component of the central thalamus. This pathway directly and indirectly innervates dopaminergic and serotonergic regions of the midbrain and dysfunction of the habenula has been linked to specific symptoms of psychiatric conditions, including depression, schizophrenia and addiction.
In a study designed to systematically explore behavioral and physiological effects of central thalamic deep brain stimulation (CT-DBS) in healthy and behaving adult nonhuman primates, we discovered that DBS of the fasciculus retroflexus (fr-DBS) induces a rapidly reversible behavioral syndrome consisting of global akinesia, conjugate slow drifting eye movements, periodic facial and upper skeletal atonia, a lack of normal response to aversive stimuli and sleep-like episodes. During fr-DBS, a consistent temporal order of the above behavioral effects was tightly correlated with enhanced cortical beta and theta band oscillations. Following fr-DBS offset, consistent stereotypical dyskinesias of the upper and lower limbs and rapid conjugate ‘non-purposeful’ eye movements were observed prior to the resumption of goal-directed behavior. These results support the hypothesis that the habenula, through the fasciculus retroflexus, regulates purposeful movement and sleep/wake activity in the primate brain as in many other vertebrate species and may provide insight into the dysfunction of the habenula, as implicated in several human psychiatric disorders.
In addition, the marked enhancement of beta-band (15-25Hz) spectral power recorded across the global ECoG array and LFPs recorded within frontal cortex and medial thalamus bear strikingly similarities to the symptoms and EEG signal abnormalities observed in human subjects suffering from basilar migraine with exception of immediate cessation upon fr-DBS offset. The robust and reproducible effects reported here may provide a mechanistic insight into this rare syndrome.
17T Diffusion Tensor Imaging Elucidates Targets of Deep Brain Stimulation
Katie Warthen1; Justin Hilliard, MD2; Kelly Foote, MD3; Christopher Butson, PhD1
1 University of Utah 2 University of Florida 3 Department of Neurosurgery
The therapeutic mechanisms of deep brain stimulation (DBS) are not well understood, partially because of the limited resolution of imaging to identify stimulation targets. To gain greater insight into this problem, we acquired pre- and postmortem imaging of a patient with Parkinson’s disease who was successfully treated using DBS. The images obtained included whole brain 3T MRI as well as 17T MRI anatomical and diffusion weighted imaging of the region around the DBS lead.
Whole brain 3T and 17T image volumes were coregistered using a visual programming computational workbench (SCIRun). Tensor reconstruction and tractography was performed using DSI Studio (http://dsi-studio.labsolver.org). The MRI images along with a simulated DBS system were evaluated in SCIRun, which was used to integrate the therapeutic volume of activation for this subject as well as anatomical areas of interest that had been segmented. The 17T image consisted of 0.19 mm3 voxels, and fiber tracts of length 10mm to 300mm were created through a reconstruction method inherent to DSI Studio. Anatomical regions were hand segmented from the 17T anatomical MRI and used to create sections of tracts corresponding to these regions of interest, also in DSI Studio. Data was then transferred to SCIRun, in which a simulated DBS system was aligned with the 17T image as well as a postmortem 3T whole brain MRI scan. Therapeutic DBS settings for the patient were then used to filter fiber tracts.
In this integrated approach we were able to identify finely spaced fiber tracts that were modulated during therapeutic DBS. Tracts activated during therapeutic DBS included those that projected to the red nucleus, subthalamic nucleus, and substantia nigra.
Knowledge of the tracts and regions modulated in each specific patient is vital to the individual success of DBS. An imaging study of this resolution has not yet been conducted for the study of DBS and could add substantially to our knowledge of target structures. There is limited knowledge about the tracts activated in DBS and the regions that they correspond to, this study helps to shed light on the neural structures affected by DBS.
Influence of Diffusion-Tensor Based Axon Orientation on the Prediction of Deep Brain Stimulation Effects
Johannes Vorwerk, PhD1; Christopher Butson, PhD
1 SCI Institute, University of Utah
Although the underlying mechanisms and the effects of deep brain stimulation (DBS) on the nervous system are not yet fully understood, computational models to predict its effects have been accepted as an accurate simulation and visualization method, and are increasingly used to postoperatively optimize stimulation protocols to target or avoid specific brain regions. A frequently used concept in this context is the volume of tissue activated (VTA). It is defined as the brain volume inside which the computational model predicts an activation (ie, initiation of an action potential) as a result of the stimulation. However, this representation can give an impression of certainty that is not necessarily justified. While the computational model itself is deterministic, its output is influenced by uncertainties inherent to the input data. One parameter that might lead to deviations between predicted and actual neural stimulation is the orientation of neuron axons. In this study the influence of uncertainty of axon directions is investigated.
Methods and Results
A realistic electric field distribution is calculated using a finite element model (FEM). Diffusion tensor imaging (DTI) data are used to determine the axon directions from deterministic tractography. For each datapoint, these are chosen as the eigenvector of the DTI tensor with the maximal eigenvalue. A reference threshold value for neuron activation is obtained by calculating the second spatial derivative in the direction of the assumed axon orientation. Next, the orientation of the axon is randomly perturbed to simulate the uncertainty in measuring the axon direction. Using the thereby generated directions, the second spatial derivative is again calculated and compared to the initial value. These data are evaluated for a range of stimulation settings. The results are both statistically and visually evaluated.
Our study presents a means to estimate the variability in predicting the effects of DBS due to uncertainties in the axon orientation as estimated from DTI data. Besides the presented direct evaluation of these effects, this approach is also applicable to estimate the stability of algorithms relying on the simulation results, e.g., for stimulation optimization, against such perturbations of the input data.