Advances in deep brain stimulation programming to guide selective neural targeting

Deep brain stimulation (DBS) has been used primarily for the treatment of movement disorders. The difference between successful and ineffective therapy often lies in stimulation parameter selection, which can be challenging to optimize. The focus of this dissertation is on specific advances in DBS programming and technology to guide and improve parameter selection for selective neural targeting. Computational modeling has been used throughout the DBS field to predict activation from stimulation, but the role of fiber orientation in these models has not been fully explored. We have found that fiber orientation influences activation thresholds, and different orientations can be selectively targeted by modifying the DBS waveform. Our results demonstrate that cathodic stimulation activates axon segments passing adjacent to the electrode, whereas anodic stimulation activates axons approaching or leaving the electrode. Accounting for fiber orientation in activation prediction models can be used to specifically target neural regions corresponding to clinical benefits. The large number of possible combinations of voltage, frequency, and pulse widths for the standard, quadripolar DBS lead makes it difficult to manually determine optimal settings, and the parameter space increases exponentially with more complex electrode geometries. We created an optimization algorithm using linear convex optimization and the Hessian matrix to maximize stimulation of a neural target and avoid stimulation outside the target. Such a programming algorithm for DBS may help reduce the time burden on iv programming physicians and patients. Further, it will be quite helpful in determining contact configurations for complex electrode designs, especially in optimizing novel, directional electrodes in DBS patients. Standard DBS technology does not provide fine stimulation resolution or stimulation steering capability in the targeting or avoidance of brain regions. We developed a novel, multiresolution DBS electrode with 864 microsized, individually controllable contacts. The novel lead, the μDBS, can stimulate at contact sizes orders of magnitude smaller than what is available in the clinic, which improves targeting of smaller diameter, therapeutic fibers. The programming flexibility offered by the μDBS may greatly improve stimulation selectivity for neural targeting. Taken together, this body of work represents a significant improvement in DBS programming that could directly impact patient care.

ADVANCES IN DEEP BRAIN STIMULATION PROGRAMMING TO GUIDE SELECTIVE NEURAL TARGETING by Daria Nesterovich Anderson A dissertation submitted to the faculty of The University of Utah in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Biomedical Engineering The University of Utah August 2019 Copyright © Daria Nesterovich Anderson 2019 All Rights Reserved The University of Utah Graduate School STATEMENT OF DISSERTATION APPROVAL The dissertation of Daria Nesterovich Anderson has been approved by the following supervisory committee members: Alan D. Dorval and by , Chair 05/03/2019 Christopher R. Butson , Member 05/03/2019 John D. Rolston , Member 05/03/2019 Robert W. Hitchcock , Member 05/03/2019 Florian Solzbacher , Member 05/03/2019 David W. Grainger the Department/College/School of Date Approved Date Approved Date Approved Date Approved Date Approved , Chair/Dean of Biomedical Engineering and by David B. Kieda, Dean of The Graduate School. ABSTRACT Deep brain stimulation (DBS) has been used primarily for the treatment of movement disorders. The difference between successful and ineffective therapy often lies in stimulation parameter selection, which can be challenging to optimize. The focus of this dissertation is on specific advances in DBS programming and technology to guide and improve parameter selection for selective neural targeting. Computational modeling has been used throughout the DBS field to predict activation from stimulation, but the role of fiber orientation in these models has not been fully explored. We have found that fiber orientation influences activation thresholds, and different orientations can be selectively targeted by modifying the DBS waveform. Our results demonstrate that cathodic stimulation activates axon segments passing adjacent to the electrode, whereas anodic stimulation activates axons approaching or leaving the electrode. Accounting for fiber orientation in activation prediction models can be used to specifically target neural regions corresponding to clinical benefits. The large number of possible combinations of voltage, frequency, and pulse widths for the standard, quadripolar DBS lead makes it difficult to manually determine optimal settings, and the parameter space increases exponentially with more complex electrode geometries. We created an optimization algorithm using linear convex optimization and the Hessian matrix to maximize stimulation of a neural target and avoid stimulation outside the target. Such a programming algorithm for DBS may help reduce the time burden on programming physicians and patients. Further, it will be quite helpful in determining contact configurations for complex electrode designs, especially in optimizing novel, directional electrodes in DBS patients. Standard DBS technology does not provide fine stimulation resolution or stimulation steering capability in the targeting or avoidance of brain regions. We developed a novel, multiresolution DBS electrode with 864 microsized, individually controllable contacts. The novel lead, the µDBS, can stimulate at contact sizes orders of magnitude smaller than what is available in the clinic, which improves targeting of smaller diameter, therapeutic fibers. The programming flexibility offered by the µDBS may greatly improve stimulation selectivity for neural targeting. Taken together, this body of work represents a significant improvement in DBS programming that could directly impact patient care. iv TABLE OF CONTENTS ABSTRACT....................................................................................................................... iii LIST OF FIGURES .......................................................................................................... vii Chapters 1 INTRODUCTION ........................................................................................................... 1 1.1 DBS Programming ............................................................................................... 2 1.2 Mechanisms of Extracellular Electrical Stimulation ........................................... 4 1.3 Directional Electrodes .......................................................................................... 5 1.4 Motivation for Work ............................................................................................ 7 1.5 Summary ............................................................................................................ 11 1.6 References .......................................................................................................... 12 2 ANODIC STIMULATION MISUNDERSTOOD: PREFERENTIAL ACTIVATION OF FIBER ORIENTATIONS WITH ANODIC WAVEFORMS IN DEEP BRAIN STIMULATION ............................................................................................................... 16 2.1 Introduction ....................................................................................................... 18 2.2 Materials and Methods ....................................................................................... 18 2.3 Results ............................................................................................................... 20 2.4 Discussion .......................................................................................................... 22 2.5 Conclusions ....................................................................................................... 25 2.6 References ......................................................................................................... 26 3 OPTIMIZED PROGRAMMING ALGORITHM FOR CYLINDRICAL AND DIRECTIONAL DEEP BRAIN STIMULATION ELECTRODES ................................ 28 3.1 Introduction ....................................................................................................... 30 3.2 Materials and Methods ...................................................................................... 31 3.3 Results ............................................................................................................... 35 3.4 Discussion .......................................................................................................... 39 3.5 Conclusions ....................................................................................................... 43 3.6 References ......................................................................................................... 43 4 THE µDBS: MULTIRESOLUTION DIRECTIONAL DEEP BRAIN STIMULATION FOR IMPROVED TARGETING OF SMALL-DIAMETER FIBERS ........................... 47 4.1 Abstract .............................................................................................................. 47 4.2 Introduction ........................................................................................................ 48 4.3 Materials and Methods ....................................................................................... 50 4.4 Results ................................................................................................................ 60 4.5 Discussion .......................................................................................................... 68 4.6 Conclusions ........................................................................................................ 71 4.7 References .......................................................................................................... 71 5 CONCLUSION .............................................................................................................. 74 5.1 Anodic Versus Cathodic Stimulation................................................................. 75 5.2 Optimization of DBS Parameters....................................................................... 76 5.3 Highly Directional Deep Brain Stimulation ...................................................... 79 5.4 DBS Programming: Moving Forward ............................................................... 81 5.5 Final Conclusions .............................................................................................. 86 5.6 References .......................................................................................................... 87 vi LIST OF FIGURES Figures 2.1. The electric potential and second derivative plotted along an axon passing tangentially to the electrode ............................................................................................. 19 2.2. Primary, secondary, and tertiary eigenvector orientations for cathodic stimulation at contact 1 of the Medtronic 3389 lead .............................................................................. 21 2.3. An unbalanced cathodic pulse causes firing of neurons oriented in the primary and secondary eigenvector directions, which represent passing axon (positive second differences, in orange) ...................................................................................................... 23 2.4. Cathode- and anode- first stimulus regimes with various charge-balancing phases of 100%, 50%, 25%, 10%, 5%, and 2.5% of the leading phase amplitude ........................... 24 2.5. Visualization of HD and IC tracts with respect to the STN (right) and lead ............. 25 3.1. Review of clinical DBS leads and emerging directional DBS technology ................ 31 3.2. Constructing the patient-specific finite element model for a DBS patient for Parkinson’s disease with the STN as the target ............................................................... 32 3.3. Exploration of the sensitivity parameter, α .............................................................. 36 3.4. Validation of algorithm-determined optimal settings ............................................... 37 3.5. Contact configuration and voltages for the Medtronic-Sapiens using different targeting constraints ......................................................................................................... 39 3.6. Summary of computation times required for all lead designs, given different targeting constraints—typically 0.1-10 s ......................................................................... 40 4.1. µDBS design and assembly ....................................................................................... 52 4.2. Design and simulation of a single contact unit on the µDBS .................................... 54 4.3. Design architecture for postprocessing ...................................................................... 55 4.4. µDBS experimental and computational setup ........................................................... 57 4.5. Programming validation of µDBS chips .................................................................... 63 4.6. Impedance and bath testing validation ....................................................................... 65 4.7. Smaller contact sizes more efficiently activate small-diameter fibers....................... 68 viii CHAPTER 1 INTRODUCTION Deep brain stimulation (DBS) is a neurostimulation therapy that has been implanted in over 160,000 patients worldwide for a number of neurological disorders (Lozano et al., 2019). DBS is primarily used to treat movement disorders, and it was first FDA approved for essential tremor and Parkinson’s disease in 1997 and 2002, respectively. The hardware setup for DBS consists of one or two stimulation electrodes implanted into the brain in a disease-specific neural target and a pulse generator implanted in the chest. The implanted pulse generator sends brief electrical pulses to the electrodes in the brain through subcutaneous leads, most commonly at frequencies around 130-185 Hz. Pulses typically last from 60-90 µs and are charge-balanced to ensure charge does not build up on the electrode metal surface and cause irreversible reactions with tissue (Kuncel & Grill, 2004; Merrill, Bikson, & Jefferys, 2005). The rapid and growing success of DBS for Parkinson’s disease and essential tremor has spurred a variety of clinical studies on its use for other disorders; however, none have reached the same level of robustness. Other disorders have received humanitarian device exemptions from the FDA, such as dystonia and obsessive compulsive disorder. Most recently, in 2018, the FDA granted approval of DBS therapy for epilepsy with the anterior nucleus of the thalamus as the target. DBS is under active investigation for a number of 2 psychiatric disorders, with promising results for Tourette’s Syndrome, treatment-resistant depression, and Alzheimer’s disease, among others. However, as more applications for DBS emerge, there is a gap in our knowledge regarding how to apply DBS to different neurological disorders, especially when therapeutic improvements are not instantly or visually apparent, such as with psychiatric disorders. Electrical stimulation is not a simple therapy to administer, and the robust success of DBS for Parkinson’s disease and essential tremor has not simply translated to other neurological disorders. In order to advance the field of DBS to applications beyond movement disorders, precise understanding of DBS programming and targeting is necessary to isolate mechanisms of stimulation. Better understanding of DBS parameter selection and programming will enable better execution of therapy while also potentially uncovering novel targets for stimulation. 1.1 DBS Programming A number of factors play a role in DBS success, namely, proper patient selection, appropriate neural target selection, accurate surgical placement, and optimal DBS settings for stimulation of the target area. After DBS surgery, all variables except for DBS settings are unchangeable, with some exceptions to surgical placement when revision surgeries might be warranted for poorly placed leads (Rolston, Englot, Starr, & Larson, 2016). Otherwise, once the DBS lead has been placed, only the stimulation parameters can be modulated to modify therapeutic outcomes, and DBS parameter selection can be the difference between successful and subpar therapy. The focus of this dissertation will explore the only changeable quality of DBS, DBS programming. Modern DBS programming is typically done through trial-and-error approaches in 3 which clinicians adjust settings based on patient responses to each setting tested (Moro et al., 2002; Rizzone et al., 2001; Volkmann, Herzog, Kopper, & Deuschl, 2002). Clinicians begin by iterating through each of the standard four contacts along a lead. Stimulation settings are first tried at low amplitudes, and then amplitudes are iteratively increased until therapeutic benefit is observed and stopped once side-effect thresholds are established. This manual, iterative process, called “monopolar review,” determines which electrode contacts might best serve as stimulation contacts for the patient and the therapeutic window, the range of amplitudes where patients receive benefit without stimulationinduced side effects, for those contacts. The contact(s) with the lowest stimulation amplitudes that result in adequate therapeutic benefit is typically chosen as the final setting. Identifying effective programming settings poses a substantial time burden on the clinician while potentially subjecting the patient to discomfort, and it is unlikely to yield optimal parameter settings since it is difficult to iterate through all possible settings. Implants for the overwhelming majority of DBS patients involve leads with four evenly spaced cylindrical contacts, which already have over 25,000 possible combinations of programming parameters such as pulse width, frequency, and voltage (Kuncel & Grill, 2004). In general, programming these leads requires approximately 18–32 hr of clinical programming time (Hunka, Suchowersky, Wood, Derwent, & Kiss, 2005) and 4-17 visits in the first year (Ondo & Bronte-Stewart, 2005). Manual programming will soon be infeasible with the emergence of novel, complex lead designs since the number of possible parameter combinations increases exponentially with the number of contacts. 4 1.2 Mechanisms of Extracellular Electrical Stimulation Deep brain stimulation has largely replaced lesion therapy in Parkinson’s disease, where tissue is removed or burned, because neurostimulation has comparable success but is instead adjustable and reversible. Stimulation parameters in Parkinson’s disease result in a therapeutic effect that appears to mimic a lesion (Agnesi, Connolly, Baker, Vitek, & Johnson, 2013; Anderson et al., 2015; Grill, Snyder, & Miocinovic, 2004); however, given that parameters are changeable, the influence of stimulation can be adjusted over time to compensate for worsening disease progression, unlike with lesion therapy. How exactly stimulation influences neural activation patterns is a current topic of investigation in DBS, and it is covered in part in this dissertation. Neurons are electrically active, and applied stimulation affects neural activity near the active electrode contacts. Axon activation in response to neural stimulation can be explained by neuronal cable theory, in which signal conduction can be modeled similar to signal transmission along infinitely long cables (McNeal, 1976). Cable theory has formed the basis of many computational experiments, especially experiments that quantify the effects of extracellular stimulation settings from DBS on nearby axons (Butson, Cooper, Henderson, & McIntyre, 2007; McIntyre & Grill, 2002). It has been shown that stimulation preferentially activates axons, rather than cell bodies in the tissue around the electrode, although some stimulation parameters may selectively activate cell bodies (McIntyre & Grill, 1999, 2000, 2002). Additionally, extracellular stimulation preferentially activates larger axons over smaller diameter axons at any distance away from the electrode (Lertmanorat & Durand, 2004). Differences in activation profiles achieved by differing DBS settings imply that DBS programming parameters act selectively on neural tissue and, 5 ultimately, that efficacy of therapy can change based on which programming settings are used. Such axon models have been used to predict the volume of tissue activated (VTA) based on patient-specific DBS settings (Butson et al., 2007). VTAs are generated by distributing individual axon models, or other cell types, at each point space around the electrode. At each axon location, the axon firing threshold is determined and used to create predictions for the spread of activation. The spread of neural activation depends on the voltage profile of the extracellular field, which is dependent on the electrode geometry, amplitude of stimulation, and duration of the pulse. The method used to model the tissue can also influence the modeled spread of activation and can vary in complexity from a simple field model in isotropic tissue to a patient-specific bioelectric field model that accounts for brain heterogeneity and anisotropy. Predictions of the VTA have guided aspects of stimulation programming and linked DBS parameters to activation of neural targets associated with clinical outcomes. 1.3 Directional Electrodes One limitation of the standard, quadripolar lead design is that there is no ability to redirect the stimulation field in cases in which the axisymmetric spread of stimulation will inadvertently activate regions that induce side effects. It has been shown that stimulation outside the target area can drive side effects, with some classic side effects being paresthesias, dyskinesias, oculomotor signs, dysarthria, impulsivity, and gait disturbance (Benabid, Chabardes, Mitrofanis, & Pollak, 2009). The appearance of side effects reduces the therapeutic window, the range of stimulation in which the patient experiences symptom 6 relief without side effects. New directional DBS lead designs use a greater number of smaller contacts that allow for directional control of stimulation spread and have been shown to widen the therapeutic window (Rossi et al., 2016). In the last few years, a number of directional leads have come onto the market, with FDA approval of the Abbott directional lead and the Boston Scientific directional lead in Europe. Both leads take the standard four-contact lead design and subdivide two of the four contacts into three smaller contacts each, creating eight total contacts. Standard cylindrical contacts are approximately 6 mm2, so the division of one contact into three smaller contacts creates three surfaces with an area of approximately 2 mm2 each. The radial subdivision of a larger contact into three smaller contacts allows for the steering of extracellular stimulation, primarily into one of three directions. Stimulation in a single direction, rather than radially, can widen the therapeutic window if stimulation is preferentially directed to the target region and away from a region that might be associated with side effects. Stimulation steering capability is important if the lead is placed off center of its target. Regardless of target, improper lead positioning relative to the target area is a common occurrence, and stimulation of unintended regions may induce psychiatric or motor side effects, leading to potentially ineffective DBS surgery (Okun et al., 2005). Even state-of-the-art stereotactic surgery for DBS has inherent positioning error, with potential deviations from the intended coordinates of up to several millimeters (Burchiel, McCartney, Lee, & Raslan, 2013). When lead placement deviation is greater than 3 mm away from the intended target, the stimulation field does not reach the area defined as optimal and outcomes are suboptimal (Ellis et al., 2008; Guridi et al., 2000). Generally, patients experience a 50% improvement in motor scores post-DBS, but outcomes vary 7 depending on the accuracy of electrode positioning (Fasano et al., 2010). An ability to precisely steer stimulation back towards the target in cases of lead placement deviation could lead to a greater likelihood of patients experiencing therapeutic improvement, and this same notion has motivated the push for the use of directional leads in the clinic. 1.4 Motivation for Work From its first introduction, DBS has become a widely researched therapy, with many advances gained through computational models, animal experiments, and human subjects research. Despite widespread use and research in DBS, mechanisms of stimulation remain unclear; this familiar critique of DBS is mentioned in nearly all research articles on DBS. With this in mind, how can we finally answer these mechanistic questions that form the basis of an invasive therapy that has already been implanted into hundreds of thousands of people? The answer to this question is multifaceted, but from the clinical perspective, our first steps are to understand which neural elements are activated by stimulation and which neural elements form the common denominator in successful DBS therapy for each neurological indication. Once proper targets of stimulation have been uncovered, as has largely been done in Parkinson’s disease, DBS programming must become both precise and reliable to activate these known targets. In an attempt to address the practical problem of applying stimulation to the appropriate targets, this dissertation focuses on how to (a) target neural populations based on fiber orientation, (b) automate the selection of contacts to maximize simulation on the region of interest, and (c) create a novel DBS lead with hundreds of microscale contacts for increased precision of neural targeting. 8 1.4.1 Selective Activation Based on Neuron Orientation Neuron orientation is heterogeneous throughout the brain, but most methods that estimate neuronal activation volumes do so assuming a homogeneous neuron orientation (Butson & McIntyre, 2005; Butson et al., 2007). We explore the influence of fiber orientation on activation trends in a systematic way, using the eigenvectors from the Hessian matrix of electric potential to describe preferential axon orientations. Grounded in the neuronal cable equation, the second difference predicts the likelihood of an axon firing based on the magnitude of the activating function (Rattay, 1986, 1999), and the Hessian matrix provides the second difference in all directions in 3D space. We are motivated in this work to fully characterize the influence of neuron orientation on activation thresholds given monopolar and bipolar stimulation configurations for a variety of stimulus waveforms. Understanding axon activation patterns based on neuron orientation may explain how DBS stimulation leads to therapeutic benefit, and understanding which local neurons are activated with specific DBS parameters is necessary to accurately target any given neural region. Historically, cathodic stimulation has been viewed as the primary driver in the activation of neural tissue in DBS (Hofmann, Ebert, Tass, & Hauptmann, 2011). However, our work has indicated that cathodic and anodic stimulation activate different populations of axons selectively based on fiber orientation. In contrast to the conventional understanding, these results indicate that anodic stimulation can also be used for therapeutic benefit in DBS. Further, differential activation of neurons based on fiber orientation around the anode could be a novel hypothesis for the reduction of stimulationinduced side effects often seen when using bipolar contact configurations (Volkmann, 9 Moro, & Pahwa, 2006). Our findings on the impact of fiber orientation on activation patterns are also a natural complement to the recent movement to use diffusion-weighted imaging (DWI) to incorporate brain anisotropy into bioelectric field modeling. Neuron fiber heterogeneity throughout the brain can be inferred from DWI, and DWI coupled with a better understanding of how to selectively activate fiber orientations with DBS programming can improve neural targeting through informed DBS parameter selection. 1.4.2 Optimization Algorithm for Automated DBS Programming Improvements to our understanding of DBS programming on axon activation may result in better targeting and clinical outcomes; however, improved programming sophistication and knowledge will, unfortunately, further complicate DBS programming. DBS programming is already time consuming and requires high levels of programming expertise. Effective DBS settings are determined through trial-and-error methods, but can almost never be “optimal” since it is impossible to manually sift through all combinations for any of the clinical DBS leads. If clinicians take into account the principles we learn on activation patterns in response to DBS parameters, DBS programming would be even more complex and nuanced than it is now, and may ultimately not lead to tangible benefits given the increased time burden on clinicians and patients. In order to incorporate specific targeting of neural fibers and nuclei while also reducing programming time, this dissertation proposes an optimization algorithm that includes targeting precision to the scale of nuclear targets and fiber tracts while fully exhausting all possible contact configurations and amplitudes in near real-time. Not only will such an optimization algorithm enable rapid, optimal DBS parameter selection in the 10 clinic, but it could also be used as a novel research tool to properly control targeting of neural structures and objectively determine DBS settings in clinical studies. By decreasing uncertainty of DBS targeting through such an optimization algorithm, we will enable fine titration of DBS therapy that may improve outcomes in challenging DBS cases and lead to the discovery of new stimulation targets in clinical research. 1.4.3 A Novel, Multiresolution Deep Brain Stimulation Electrode The DBS field has taken steps to achieve more precise targeting through DBS programming by revising clinical lead designs so that some electrodes can directionally steer stimulation. Even though directional leads that have been approved for clinical use have reduced contact size, clinical directional electrodes still produce massive stimulation fields, much larger than the microscale of the neurons they activate. The limited precision in clinical DBS devices motivates a novel lead design, the µDBS, with hundreds of microscale contacts that can each be individually controlled. The device we have built and validated in this dissertation offers multiresolution contacts, with the ability to stimulate through contacts that range in size from 150 x 150 µm to the size (or larger) of contacts already found in the clinic. Improving DBS lead technology to have more contact configuration combinations will also translate to greater flexibility during DBS programming. Because DBS parameter selection is the only changeable factor of DBS once the lead has been implanted, having more programming options with a multiresolution device could increase the likelihood of successful therapy. The µDBS, in itself, is a novel technology that does not resemble any DBS electrode in clinical use or those explored through other directional lead studies. The µDBS 11 incorporates 864 contacts, orders of magnitude greater than modern DBS technology. Novel directional leads have been developed and explored, but they largely build upon already FDA-approved technology, in which each contact corresponds to an additional wire. However, the maximum number of wires possible, on the order of tens of wires, is physically limited by what could pass subcutaneously from the lead in the brain to the implanted pulse generator in the chest. Our technology takes advantage of digital logic to multiplex each contact so that all 864 contacts can be controlled individually with only 12 input wires. The µDBS offers unprecedented control of stimulation profiles with the ability to set each contact to one of seven active voltage states, resulting in 10780 total combinations of contact configurations. Additionally, the microarray configuration of the µDBS is novel and can produce stimulation fields with a resolution closer to the scale of the neuron fibers being targeted. Such a lead design has multiresolution contacts that can stimulate with small fields locally, but can also stimulate at contact sizes that match what is currently used in the clinic for a larger field of activation. Using such a device would be impossible if it were to be manually programmed, but in combination with the optimization algorithm and neural targeting work discussed in this dissertation, we believe it is the opportune time to push DBS technology and programming forward to enable more selective neural targeting. 1.5 Summary The three projects discussed in this dissertation take three approaches to improving DBS programming, through improvements to our understanding of DBS targeting based on fiber orientation, a near real-time optimization algorithm, and hardware advancements through a novel DBS electrode with hundreds of microscale contacts. 12 The effectiveness of DBS varies across patients, and success depends largely on proper stimulation of the target area. Although ideal surgical targets are still debated across some disorders, there remains the practical challenge of effectively applying stimulation to the intended target once the DBS lead has been implanted. With better DBS programming, nonresponders may become responders, and adequate therapy may become optimal therapy. As the only adjustable factor after surgery, successful DBS programming is directly responsible for positive clinical outcomes for a wide range of neurological indications. Improved understanding of fiber-level targeting, the optimized programming algorithm, and the multiresolution novel electrode established in this dissertation will each contribute to improving DBS programming capabilities. Additionally, the unprecedented fieldshaping flexibility offered by the µDBS will be useful as a transformative research tool to finely titrate therapy, thus potentially growing our understanding of fiber-level targeting and aiding in the search for better stimulation targets in DBS. 1.6 References Agnesi, F., Connolly, A. T., Baker, K. B., Vitek, J. L., & Johnson, M. D. (2013). Deep brain stimulation imposes complex informational lesions. PloS One, 8(8), e74462. Anderson, C. J., Sheppard, D. T., Huynh, R., Anderson, D. N., Polar, C. A., & Dorval, A. D. (2015). Subthalamic deep brain stimulation reduces pathological information transmission to the thalamus in a rat model of Parkinsonism. Frontiers in Neural Circuits, 9, 31. Benabid, A. L., Chabardes, S., Mitrofanis, J., & Pollak, P. (2009). Deep brain stimulation of the subthalamic nucleus for the treatment of Parkinson’s disease. The Lancet Neurology, 8(1), 67–81. https://doi.org/10.1016/S1474-4422(08)70291-6 Burchiel, K. J., McCartney, S., Lee, A., & Raslan, A. M. (2013). Accuracy of deep brain stimulation electrode placement using intraoperative computed tomography 13 without microelectrode recording. Journal of Neurosurgery, 119(2), 301–306. https://doi.org/10.3171/2013.4.JNS122324 Butson, C. R., Cooper, S. E., Henderson, J. M., & McIntyre, C. C. (2007). Patient-specific analysis of the volume of tissue activated during deep brain stimulation. NeuroImage, 34(2), 661–670. https://doi.org/10.1016/j.neuroimage.2006.09.034 Butson, C. R., & McIntyre, C. C. (2005). Tissue and electrode capacitance reduce neural activation volumes during deep brain stimulation. Clinical Neurophysiology : Official Journal of the International Federation of Clinical Neurophysiology, 116(10), https://doi.org/10.1016/j.clinph.2005.06.023 Ellis, T.-M., Foote, K. D., Fernandez, H. H., Sudhyadhom, A., Rodriguez, R. L., Zeilman, P., … Okun, M. S. (2008). Reoperation for suboptimal outcomes after deep brain stimulation surgery. Neurosurgery, 63(4). Retrieved from http://journals.lww.com/ neurosurgery/Fulltext/2008/10000/REOPERATION_FOR_SUBOPTIMAL_OUT COMES_AFTER_DEEP.26.aspx Fasano, A., Romito, L. M., Daniele, A., Piano, C., Zinno, M., Bentivoglio, A. R., & Albanese, A. (2010). Motor and cognitive outcome in patients with Parkinson’s disease 8 years after subthalamic implants. Brain, 133(9), 2664–2676. https://doi.org/10.1093/brain/awq221 Grill, W. M., Snyder, A. N., & Miocinovic, S. (2004). Deep brain stimulation creates an informational lesion of the stimulated nucleus. Neuroreport, 15(7), 1137–1140. Guridi, J., Rodriguez-Oroz, M., Lozano, A., Moro, E., Albanese, A., Nuttin, B., … Obeso, J. (2000). Targeting the basal ganglia for deep brain stimulation in Parkinson’s disease. Neurology, 55(12 Suppl 6), S21–S28. Retrieved from PubMed. (11188971) Hofmann, L., Ebert, M., Tass, P., & Hauptmann, C. (2011). Modified pulse shapes for effective neural stimulation. Frontiers in Neuroengineering, 4, 9. https://doi.org/10.3389/fneng.2011.00009 Hunka, K., Suchowersky, O., Wood, S., Derwent, L., & Kiss, Z. H. T. (2005). Nursing time to program and assess deep brain stimulators in movement disorder patients. Journal of Neuroscience Nursing, 37(4). Retrieved from http://journals.lww.com/ jnnonline/Fulltext/2005/08000/Nursing_Time_to_Program_and_Assess_Deep_Br ain.6.aspx Kuncel, A. M., & Grill, W. M. (2004). Selection of stimulus parameters for deep brain stimulation. Clinical Neurophysiology, 115(11), 2431–2441. https://doi.org/ 10.1016/j.clinph.2004.05.031 Lertmanorat, Z., & Durand, D. M. (2004). Extracellular voltage profile for reversing the 14 recruitment order of peripheral nerve stimulation: A simulation study. Journal of Neural Engineering, 1(4), 202–211. https://doi.org/10.1088/1741-2560/1/4/003 Lozano, A. M., Lipsman, N., Bergman, H., Brown, P., Chabardes, S., Chang, J. W., … Krauss, J. K. (2019). Deep brain stimulation: Current challenges and future directions. Nature Reviews Neurology, 15(3), 148–160. https://doi.org/10.1038/ s41582-018-0128-2 McIntyre, C. C., & Grill, W. M. (1999). Excitation of central nervous system neurons by nonuniform electric fields. Biophysical Journal, 76(2), 878–888. https://doi.org/ 10.1016/S0006-3495(99)77251-6 McIntyre, C. C., & Grill, W. M. (2000). Selective microstimulation of central nervous system neurons. Annals of Biomedical Engineering, 28(3), 219–233. https://doi.org/10.1114/1.262 McIntyre, C. C., & Grill, W. M. (2002). Extracellular stimulation of central neurons: influence of stimulus waveform and frequency on neuronal output. Journal of Neurophysiology, 88(4), 1592. McNeal, D. R. (1976). Analysis of a model for excitation of myelinated nerve. IEEE Transactions on Biomedical Engineering, BME-23(4), 329–337. https://doi.org/ 10.1109/TBME.1976.324593 Merrill, D. R., Bikson, M., & Jefferys, J. G. (2005). Electrical stimulation of excitable tissue: design of efficacious and safe protocols. Journal of Neuroscience Methods, 141(2), 171–198. Moro, E., Esselink, R. J. A., Xie, J., Hommel, M., Benabid, A. L., & Pollak, P. (2002). The impact on Parkinson’s disease of electrical parameter settings in STN stimulation. Neurology, 59(5), 706–713. https://doi.org/10.1212/WNL.59.5.706 Okun, M. S., Tagliati, M., Pourfar, M., Fernandez, H. H., Rodriguez, R. L., Alterman, R. L., & Foote, K. D. (2005). Management of referred deep brain stimulation failures: A retrospective analysis from 2 movement disorders centers. Archives of Neurology, 62(8), 1250–1255. Ondo, W. G., & Bronte-Stewart, H. (2005). The North American Survey of placement and adjustment strategies for deep brain stimulation. Stereotactic and Functional Neurosurgery, 83(4), 142–147. Rattay, F. (1986). Analysis of models for external stimulation of axons. IEEE Transactions on Biomedical Engineering, BME-33(10), 974–977. https://doi.org/10.1109/ TBME.1986.325670 Rattay, F. (1999). The basic mechanism for the electrical stimulation of the nervous system. 15 Neuroscience, 89(2), 335–346. https://doi.org/10.1016/S0306-4522(98)00330-3 Rizzone, M., Lanotte, M., Bergamasco, B., Tavella, A., Torre, E., Faccani, G., … Lopiano, L. (2001). Deep brain stimulation of the subthalamic nucleus in Parkinson’s disease: Effects of variation in stimulation parameters. Journal of Neurology, Neurosurgery, and Psychiatry, 71(2), 215–219. https://doi.org/10.1136/ jnnp.71.2.215 Rolston, J. D., Englot, D. J., Starr, P. A., & Larson, P. S. (2016). An unexpectedly high rate of revisions and removals in deep brain stimulation surgery: Analysis of multiple databases. Parkinsonism & Related Disorders, 33, 72–77. https://doi.org/10.1016/j.parkreldis.2016.09.014 Rossi, P. J., Gunduz, A., Judy, J., Wilson, L., Machado, A., Giordano, J. J., … Okun, M. S. (2016). Proceedings of the Third Annual Deep Brain Stimulation Think Tank: A review of emerging issues and technologies. Frontiers in Neuroscience, 10, 119. https://doi.org/10.3389/fnins.2016.00119 Volkmann, J., Herzog, J., Kopper, F., & Deuschl, G. (2002). Introduction to the programming of deep brain stimulators. Movement Disorders, 17(S3), S181–S187. https://doi.org/10.1002/mds.10162 Volkmann, J., Moro, E., & Pahwa, R. (2006). Basic algorithms for the programming of deep brain stimulation in Parkinson’s disease. Movement Disorders, 21(S14), S284–S289. https://doi.org/10.1002/mds.20961 CHAPTER 2 ANODIC STIMULATION MISUNDERSTOOD: PREFERENTIAL ACTIVATION OF FIBER ORIENTATIONS WITH ANODIC WAVEFORMS IN DEEP BRAIN STIMULATION Reprinted under the Creative Commons license from the Journal of Neural Engineering. Authors: Daria Nesterovich Anderson, Gordon Duffley, Johannes Vorwerk, Alan D Dorval, and Christopher R Butson. Journal Citation: Daria Nesterovich Anderson et al 2019 J. Neural Eng. 16 016026 DOI: https://doi.org/10.1088/1741-2552/aae590 17 18 19 20 21 22 23 24 25 26 27 CHAPTER 3 OPTIMIZED PROGRAMMING ALGORITHM FOR CYLINDRICAL AND DIRECTIONAL DEEP BRAIN STIMULATION ELECTRODES Reprinted under the Creative Commons license from the Journal of Neural Engineering. Authors: Daria Nesterovich Anderson, Braxton Osting, Johannes Vorwerk, Alan D Dorval, and Christopher R Butson. Journal Citation: Daria Nesterovich Anderson et al 2018 J. Neural Eng. 15 026005 DOI: https://doi.org/10.1088/1741-2552/aaa14b 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 CHAPTER 4 THE µDBS: MULTIRESOLUTION DIRECTIONAL DEEP BRAIN STIMULATION FOR IMPROVED TARGETING OF SMALL-DIAMETER FIBERS 4.1 Abstract Directional deep brain stimulation (DBS) leads have recently been approved and used in patients, and growing evidence suggests that directional contacts can increase the therapeutic window by redirecting stimulation to the target region while avoiding sideeffect-inducing regions. In this chapter, we outline the design, fabrication, and testing of a novel directional DBS lead, the µDBS, which utilizes microscale contacts to increase the spatial resolution of stimulation steering and improve the selectivity in targeting small diameter fibers. We outline the steps of fabrication of the µDBS, from an integrated circuit design to postprocessing and validation testing. We tested the onboard digital circuitry for programming fidelity, characterized impedance for a variety of electrode sizes, and demonstrated functionality in a saline bath. In a computational experiment, we determined that reduced electrode sizes focus the stimulation effect on small, nearby fibers. We determined distances at which larger fibers were activated by the same voltage as small, nearby fibers through standard electrode sizes and then found activation thresholds for these same model neurons with our variable electrode sizes. Smaller electrode sizes allow 48 for a relative decrease in small-diameter axon thresholds compared to thresholds of largediameter fibers, demonstrating focusing of the stimulation effect within small and, possibly, therapeutic fibers. This principle of selectivity could be useful in further widening the window of therapy. The µDBS offers a unique, multiresolution design in which any combination of microscale contacts can be used together to function as electrodes of various shapes and sizes. Multiscale electrodes could be useful in selective neural targeting for established neurological targets and in exploring novel treatment targets for new neurological indications. 4.2 Introduction Deep brain stimulation (DBS) is a widely accepted therapy for several movement disorders and an emerging therapy for psychiatric disorders and additional movement disorders. From its first FDA approval for essential tremor in 1997, the physical design of DBS leads has remained largely unchanged (Eisinger, Cernera, Gittis, Gunduz, & Okun, 2019). The classic lead design is defined by a cylindrical shaft with four cylindrical electrode contacts. In this chapter, we present a novel neurostimulation device that assembles multiresolution electrodes from microscale contacts to enable fine control of the stimulation volume and an improved capability to target small-diameter fibers. In recent years, the FDA has approved more lead designs from major neuromodulation companies; however, these leads differ minimally from the classic quadripolar lead design. Moderate advances to the classic lead design involve contacts capable of directionally focusing stimulation, typically by having two of the four contacts subdivided into three smaller contacts each. These smaller, directional contacts allow for 49 directional steering of the activation field to, ideally, activate the target structure avoiding side-effect-inducing regions that might reduce the window of therapy. Directional stimulation has already been clinically demonstrated to widen the therapeutic window by steering stimulation away from regions that may be responsible for inducing side effects (Steigerwald, Müller, Johannes, Matthies, & Volkmann, 2016). Other experimental lead designs have further subdivided contacts to allow for finer directional control and have shown promising results at widening the therapeutic window (Contarino et al., 2014; Pollo et al., 2014). However, the fundamental limitation in repeatedly subdividing contacts is enclosing enough wires for each contact within the lead shaft without increasing the width of the lead. In this chapter, we propose a novel directional DBS device, the µDBS, with hundreds of individually controllable contacts capable of stimulation and recording. Using onboard circuitry, the lead can stimulate using any combination of contacts at 7 independent voltage states with only 12 input wires. Multiresolution electrode sizes and complex monopolar and bipolar configurations are achievable by grouping contacts according to the desired stimulation bus lines. Such flexibility enables electrodes to scale in size from the ~6.0 mm2 of the classic clinical electrode down to the ~0.02 mm2 of a single µDBS contact. The chapter outlines the design steps, fabrication, and bench testing of this novel, multiresolution DBS device. We aim to push the field toward DBS leads with the capability of stimulating through variously sized electrodes composed of contacts that are orders of magnitude smaller than currently available in the clinic. In many instances, the side-effect-inducing regions comprise larger fibers than those most associated with therapeutic benefit 50 (Chaturvedi, Butson, Lempka, Cooper, & McIntyre, 2010; Lang, Lozano, Ashby, Kumar, & Kim, 1999). In this chapter, we provide additional computational evidence that smaller electrodes more efficiently activate small-diameter fibers over large-diameter fibers (C.J. Anderson et al., Anderson, Pulst, Butson, & Dorval, 2019). Given this finding, using smaller contacts can be even more advantageous than simply improving steering resolution and flexibility; smaller contacts may also widen the therapeutic window by preferentially activating smaller, therapeutic fibers over larger, side-effect-inducing fibers. The present work supports that multiresolution stimulation devices can substantially improve neuromodulation efficiency and selectivity, and demonstrates the practicality of building one such device, the µDBS, as part of the next generation of neuromodulation therapy. 4.3 Materials and Methods We designed a novel DBS lead, the µDBS, to resemble a microelectrode array applied to deep brain stimulation leads that has a similar scale to the clinical lead, with a width of approximately 1.27 mm, but is composed of 864 microscale contacts instead of 4 large contacts. This chapter expands upon our first iteration of the µDBS (Willsie & Dorval, 2015b) with improvements to the fabrication process, larger stimulation contacts, and increased stimulation flexibility via the incorporation of seven stimulation bus lines. The novel lead is fabricated using silicon wafer-based technology, and its onboard digital circuitry allows for full control to open or close any combination of the 864 contacts using only 12 input wires. The small contact size on the µDBS — 0.0225 mm2 compared to the 6 mm2 for the clinical electrode (Lanotte et al., 2002) — allows for the µDBS to have 864 total contacts and still match the overall size of the clinical lead, having a width of 1.27 51 mm (Figure 4.1A). A complete lead is assembled from four silicon chips consisting of 216 contacts each; two pairs of flat chips are assembled front-to-back, and each pair of chips slides past the other to form a plus-shaped structure (Figure 4.1B). Through our redesign of DBS lead technology, the µDBS is the first DBS lead of similar size to the clinical leads capable of stimulating through multiresolution electrodes made up of hundreds of microscale contacts. 4.3.1. Design and Fabrication In order to achieve hundreds of individually controllable contacts, the µDBS must have on-board digital circuitry, unlike modern DBS leads that have a single wire to power each conductive contact. Each µDBS contact is programmable to eight possible states using three-bit digital logic (23 = 8). Seven of the eight states tie the contact to bus lines that can be used to stimulate or record, and the last state is reserved as an unconnected, floating state. Each bus line active state is independent from the others, which allows for flexibility in stimulation, frequency, pulse width, and waveform shape for each electrode used on the lead. Having these multiple independent sources allows for greater spatial and temporal flexibility in stimulation shaping since electrodes could take on various shapes, be used in complex multipolar and bipolar configurations, and stimulate at different programming settings and waveforms. Programming the device requires transmitting a serial program of three-bit “words,” where each word determines the bus line to which the contact will be tied. Each contact stores three bits of information across a shift register (serial cascade of three flip flops) and advances each bit during the falling phase of a clock signal until all contacts 52 Figure 4.1. µDBS design and assembly. A. Clinical deep brain stimulation electrode (left) with four contacts, and the μDBS (right) with hundreds of contacts. B. The μDBS electrode is assembled from four total flat chips, with two flat chips paired back to back. The paired chips are assembled together to form a ‘+’ shape when viewed from above. 53 have been programmed to the intended state (Figure 4.2A). We tested whether contact states could be theoretically programmed using three-bit digital logic through the simulation of a single contact circuit prior to fabrication by X-FAB (Figure 4.2B). Given the presence of onboard circuitry and the serial nature of the circuit design, all contacts are controllable with a minimal number of wires using five inputs (input program, clock, power, ground, power switch) and up to seven different bus line inputs. A layout design was made in Cadence Virtuoso using the XC06 (0.6 µm) technology package from X-FAB foundry (X-FAB, Erfurt, Germany). Circuitry for a single contact unit can be found in the left panel of Figure 4.2C, and the VLSI design was validated using the Cadence ADE XL package. Images of the single contact postfabrication and postprocessing can be found in Figure 4.2C, in the middle and right panels, respectively. Circuitry associated with one contact resides within a 165 µm x 165 µm patch, enabling a total contact size of 150 µm x 150 µm with 15 µm spacing between contacts. The primary fabrication of the design was performed by X-FAB, and postprocessing fabrication work was performed in the Utah Nanofab Cleanroom at the University of Utah. The foundry fabricated chips include three small subcontact pads per contact unit that underwent further processing to be linked into a single contact (see Figure 4.2C). Additionally, the unprocessed contact pads used Al contacts (0.5% Cu) which are not biocompatible. Chips from the foundry were sputtered with a titanium adhesion layer (~30 nm), followed by ~270 nm of gold, which is nontoxic and nonreactive to tissue (Merrill, Bikson, & Jefferys, 2005). Afterwards, the chips underwent photolithography and patterning of negative photoresist (AZ nLoF 2020) in the shape of the desired contact size, at 150 µm x 150 µm (Figure 4.3). We exposed the patterned chips to a gold etch (8% I2, 54 Figure 4.2. Design and simulation of a single contact unit on the µDBS. A. Single contact circuit diagram with three-bit digital logic for the gating of seven bus lines. B. Simulation demonstrating programming of different bus lines on a single contact in Cadence ADE XL. With the example bit stream, 011101000, we demonstrate programming the flip flop states at the falling phase of the clock signal. C. Integrated circuit layout design of a single contact used in the simulation (left), postfabrication view of the VLSI design (middle), and view of contact after gold application in postprocessing (right). Note that for any moment in time, at most 1 of the 7 bus lines can be connected to a contact (large, bright gold square at right) through one of its 3 subcontact conduits (small, dull white squares shown in the middle panel). 55 Figure 4.3. Design architecture for µDBS postprocessing. Fabricated chips (0) undergo gold deposition (1) and are covered with AZ nLoF 2020 negative photoresist (2). Photoresist is exposed to UV light according to the desired contact layout through a photolithography mask and regions of photoresist not exposed to light are removed (3). Gold and titanium layers are etched away from regions not covered by photoresist to reveal gold contacts (4). Remaining photoresist is washed off (5) and the chips are diced to the appropriate size (6). Connection pads are wirebonded to test PCBs (7) to enable device programming and functionality testing. Silicone was used to insulate noncontact regions from water exposure during the validation experiments. 56 21% KI, 71% DI) and a titanium etch (20:1:1 DI:HF:H2O2) to clear the titanium/gold layer from noncontact areas. Afterwards, we diced the test structures placed during fabrication along the edge of the chip to match the width of the clinical electrode sizing of 1.27 mm using a diamond blade saw 70 µm in width. Following the postprocessing and cutting of the device, we mounted and wirebonded the chips using aluminum wire onto a customprinted PCB to enable µDBS programming through a computer. An interface piece of silicon with gold traces was used to facilitate wirebonding from the µDBS chip to the PCB. The design and fabrication steps discussed in this section outline novel technology necessary to build DBS leads capable of multiresolution electrode sizes for unprecedented stimulation flexibility. The onboard circuitry and three-bit programming logic enables each contact to be individually controllable, and full functionality of the device can be achieved through only twelve wires. In the following section, we demonstrate functionality of the µDBS design through a series of programming, impedance, and stimulation bench tests. 4.3.2 Validation We verified the ability to program the µDBS through a series of bench tests. In the scope of this chapter, we will examine the functionality of the µDBS by fabricating and testing single flat chips that have 216 contacts each. The first instance of testing is to determine the accuracy of programming of an intended contact configuration (Figure 4.4A). We measured the success rate of programming each electrode state during the falling phase of the clock cycle. A numerically randomized series of 648 binary numbers (i.e., ones and zeros) was generated to program three bits on each of the 216 contacts using an Arduino programming setup, repeated five times per chip at six clock speeds. Programming 57 Figure 4.4. µDBS experimental and computational setup. A. Experimental setup for programming requires only input/output information from the µDBS chip interfaced with the Arduino and computer. B. Impedance testing requires a potentiostat connected to one bus line of the µDBS, a Pt counter wire, and an Ag/AgCl reference wire in a saline bath. C. Bath testing uses a CNC machine to move a voltage probe in the saline bath around the µDBS. Voltage recordings run through a peak detection circuit and to the Arduino for recording. D. A lead-in-the-box model was used to simulate the voltage spread; multicompartment models were used to measure the effects of contact size on activation for 2.0 µm, 5.7 µm, and 10.0 µm diameter axons. 58 errors were quantified on 10 chips by comparing the fidelity of the bit program after it had passed through the chip to the series of bits that were programmed into the chip. Additionally, we measured changes in impedances in a saline bath based on the number of contacts recruited. Increasing the number of contacts recruited to a single bus line increases the surface area of the effective electrode. The total electrode impedance was expected to vary with approximate inverse proportionality to the electrode surface area. To test proper contact recruitment, we prepared a saline solution (0.1 w/v% NaCl) to match the conductivity of brain tissue at 0.2 S/m to simulate the expected impedance of the electrode when exposed to a biological environment. Impedances were measured on a commercial electrochemical test system (Gamry Instruments PC4 Potentiostat, Warminster, PA) across a Ag/AgCl reference electrode, a Pt wire counter electrode, and active contacts of the µDBS as the working electrode (Figure 4.4B). Impedances were quantified over a frequency range of 10 Hz to 10 kHz with a sinusoidal input voltage of 10 mV. The number of active contacts constituting the active µDBS electrode varied from 1 to 108, and each configuration was repeated three times for each of three chips. Finally, we experimentally measured the stimulation field produced by the µDBS for two electrode configurations using an Ag/AgCl voltage probe manipulated by a computer numerical control (CNC) machine in a saline bath that matched the conductance of neural tissue (Figure 4.4C). The CNC machine moved the probe at a 0.5 mm resolution in a 20 mm x 10 mm grid in front of the µDBS chip in the saline solution. The purpose of this experiment was to verify that stimulation can be done with simultaneous bus lines at different settings. The chip was functionally split in two, with 48 contacts on one half of the chip tied to bus line A and another 48 contacts on the other half tied to bus line B. For 59 one condition, bus line A was 1.5 V and bus line B was 3.0 V with 100 µs, charge-balanced pulses; for a second condition, the bus lines were swapped. Stimulation profiles were collected for one chip in three trials for both conditions to evaluate whether there was a directional shift in the measured field, consistent with which contacts had been assigned to which bus lines. 4.3.3 Computational Model To support the need for a multiresolution device with contacts as small as 150 µm x 150 µm, we ran computational axon models to simulate the influence of activation fields from the µDBS on neurons for varying electrode sizes. Each vertical column on the µDBS comprises 36 contacts. We ran bioelectric field solutions in SCIRun 4.7 (Scientific Computing and Imaging (SCI), Institute, University of Utah, Salt Lake City, UT) for 1-36 adjacent contacts within a column set to -1 V with the surrounding box set to 0 V. These configurations resulted in electrode sizes from 150 µm x 150 µm to 150 µm x 6 mm. We implemented a high-resolution submesh with 0.1 mm spacing around the electrode, as we have previously implemented (D.N. Anderson, Ostin, Vorverk, Dorval, & Butson, 2018; D.N. Anderson, Duffley, Vorverk, Dorval, & Butson, 2019), and we set tissue conductivity to 0.2 S/m (Figure 4.4D). Noncontact regions of the µDBS were modeled as ideal insulators, and the contacts were modeled as ideal conductors. Neurons of various diameters — 2.0 µm, 5.7 µm, and 10.0 µm — were placed parallel to the lead in 0.1 mm increments, from 0.1 to 10 mm away. The vertical axonal orientation of neurons was chosen to match that of the active electrode on the µDBS, to explore the effects of electrode size on neuron activation patterns. Neuron simulations were run in NEURON 7.4 using the 60 MRG neuron model (McIntyre, Richardson, & Grill, 2002), and thresholds were identified for a 90 µs charge-balanced pulse at ~0.01 V resolution to determine the role of electrode size on neural selectivity based on fiber diameter. 4.4 Results We conducted bench testing to evaluate the functionality of the fabricated and postprocessed µDBS chips. We determined whether chips met our design specifications, as well as whether contacts could be assigned bus lines and recruited into larger, multiresolution electrodes through programming and electrical testing. 4.4.1 Design Verification A series of sixteen chips were slated for postprocessing and subsequent testing. Of those, six were irrevocably damaged, primarily at the wirebonding post-processing step. Table 4.1 summarizes the design specifications and results retrieved for the ten surviving chips. Final chip widths were ~1.29 mm, and well within 5% tolerance of 1.27 mm design specification used to match the clinical lead. Most devices (7/10) met our form-factor and bus line acceptance criteria. In the other devices (3/10), wirebonding failed to connect all seven bus lines; but note that these chips could still be used with somewhat reduced flexibility through their 4–6 functioning bus lines. We attempted gold contact patterning on eight of the ten chips, and they all met acceptance criteria. On average, the gold contact widths and heights measured slightly smaller than designed, possibly because of chemical undercutting from the gold and titanium etch during photolithography. To accommodate this undercut in future iterations, we will simply enlarge the contacts in the 61 Table 4.1: Design Verification to Determine Whether Devices Met Design Specifications. 62 photolithography mask. In summary, gold contact patterning was universally successful, and the majority of chips met all acceptance criteria; for chips that did not meet acceptance criteria, wirebonding was the most common failure point. 4.4.2 Programming Testing We tested 10 µDBS chips for programming fidelity (Figure 4.5): chips must be programmed properly in order to stimulate properly. We randomized a series of (216 contacts × 3 bits/contact =) 648 bits for each programming trial, to give a diverse range of maximally disordered configurations. The minimal programming duration for a single chip was limited to ~2.7 s by the maximal clock rate of the Arduino device we used for programming. Since a complete µDBS lead comprises four flat chips, programming an entire lead with this device would take ~10.8 s. Programming times of 2.7–14.3 s at six different clock speeds were tested five times each, for a total of thirty programming sessions per chip. Some chips exhibited no errors in any session, and there was no significant relationship between programming time and error rate (Figure 4.5C, p = 0.97, ANOVA). Thus, chips could likely be programmed in much less than 2.7 s, given appropriately high-clock rate controllers. Figure 4.5A reports that three chips (#’s 8-10) did not have any errors regardless of settings, and four other (#’s 4-7) had relatively rare and/or constrained errors. In one of its thirty trials, chip #7 encountered one deletion error, where one missed bit initiated a cascade effect resulting in many improperly set contacts in that one trial. However, most of the chip errors were programming mutations, where one bit was toggled inappropriately. Because individual contacts are so small and operate in parallel with the other contacts 63 Figure 4.5. Programming validation of µDBS chips. A. Characterization of contact errors for 10 µDBS chips across 30 trials each, with incorrectly programmed contacts denoted as a dot (top), and as percentage distribution intervals (95/75/50/25/5) of contact errors (middle) and bit errors (bottom). Programming errors are largely chip-specific, with three chips (#’s 8-10) not displaying any programming errors. The location of programming errors appeared to cluster on nearby contacts. B. Heatmaps of programming contact errors for each chip demonstrate that errors cluster on similar regions for each chip. C. There was no significant trend that programming time affected programming error rate, indicating that chip-specific programming errors are independent of the clock rate. 64 composing a shared electrode, lone mutation errors would not substantially impact functionality. Across all the randomized trials, >95% of errors arose from a mutation toggling one bit from a low to a high state, which indicates possible crosstalk between programming connections. Contact errors are summarized in heat maps on each chip in panel Figure 4.5B. Most errors on any given chip recurred at similar locations — denoted with yellow-to-red coloring — which may indicate circuit damage that could have occurred during handling. 4.4.3 Electrode Testing The programming tests verify that each contact could be programmed to a bus line, but verifying that the contacts successfully link to the intended bus line requires electrical of the electrodes. For electrical testing, we submerged µDBS chips in a saline bath and programmed them. In separate experiments, we measured the effective impedance of electrodes built from various numbers of contacts, and assessed the spatial voltage profile generated by two separate electrodes driven by two separate bus lines on the same chip. Three chips were submerged into saline solution and connected to a computer for programming and a potentiostat for impedance testing. Impedances were recorded on each chip for electrodes programmed to range from 1 to 108 contacts, for a range of frequencies. Each recording was repeated three times, and the resulting impedance magnitude and phase spectra are shown in Figure 4.6A. Consistent with studies of other electrodes, impedances were higher at lower frequencies due to capacitance at the electrode-tissue interface As expected, impedance was inversely proportional to the electrode surface area (i.e., the number of active contacts) supporting that the contacts were properly 65 Figure 4.6. Impedance and bath testing validation. A. Magnitude and phase of impedance for one through 108 contacts activated. Impedance decreases at higher frequencies. B. Impedance is inversely proportional to surface area. Average impedance for a single contact is 178.4 kΩ with a trend towards increasing with the number of active contacts. C. Bath testing demonstrates a directional shift in the normalized voltage field depending on the relative amplitudes of the electrode voltages (p<0.00001, two-sample t test). 66 programmed and linked to the appropriate bus line. Figure 4.6B summarizes impedance values at 1 kHz, the frequency most commonly used to report impedances of clinical DBS devices. At 1kHz, a single contact has an impedance of ~180 kΩ, yielding an effective electrode impedance of ~180 kΩ divided by the number of constitutive contacts. Thus, an electrode comprising 90 contacts has a surface area of ~2.0 mm2 and an impedance of ~2.0 kΩ, matching (to within a few percent) the corresponding parameters of clinically approved directional electrodes (Butson, Maks, & McIntyre, 2006; Rebelo et al., 2018). Finally, the field-testing experiment from Figure 4.4C was performed on one chip to validate stimulation fields generated from two simultaneously active bus lines (Figure 4.6C). Two groups of 48 contacts on each half of the chip were tied to one of the two bus lines. In the two conditions tested, either bus line A was greater than bus line B, or vice versa. The voltage probe, traveling in a 20 mm × 10 mm grid in front of the stimulation electrodes, recorded a shift in the peak voltage based on which side of the µDBS lead was tied to the larger amplitude bus line (p<0.00001, two-sample t test). Although our experimental configuration did not allow for a comprehensive mapping of the voltage field, these results demonstrate that the distinct electrodes on opposite sides of a µDBS chip are capable of properly stimulating with separate voltage signals. 4.4.4 Computational Experiment Our final experiment demonstrates a possible advantage to a multiresolution device like the µDBS. Our group has previously shown that the µDBS is capable of fine directional control of the stimulation field (Willsie & Dorval, 2015a), but through this computational experiment, we demonstrate that there is increased selectivity for small-diameter fibers 67 with smaller, directional contacts. We modeled individual axons in NEURON responding to voltage fields generated via µDBS electrodes as simulated in SCIRun. Initial electrodes were modeled as 9 vertically stacked contacts, or a 1.47 mm electrode height, to approximate the extent of standard cylindrical DBS electrodes. Modeling axons of three diameters — 2.0, 5.7, and 10.0 µm — running parallel to the electrode, we positioned each fiber to its threshold distance at which a -1 V stimulation elicited an action potential. We then varied the number of active contacts within the electrode from 1 to 36, and determined the threshold voltage at which each axon fired (Figure 4.7). Smaller electrodes preferentially excited smaller axons. In the extreme case of a single-contact electrode (i.e., 150 µm), 2.0 µm diameter fibers were activated at 66% and 58% of the 5.7 µm and 10.0 µm fiber thresholds, respectively. Conversely, larger electrodes preferentially excited larger axons. In the extreme case of a 36-contact electrode (i.e., ~6 mm), 10.0 µm fibers were activated at fibers were activated at approximately 50% and 35% of the 5.7 µm and 2.0 µm fiber thresholds, respectively. Thus, the ability to use smaller electrodes ay open therapeutic window by increasing the activation of small, nearby, and likely therapeutic fibers, while decreasing the activation of large, distant, and likely sideeffect-inducing fibers. 4.5 Discussion This chapter discusses fabrication and testing of the µDBS device, a novel DBS lead with hundreds of individually controllable contacts. Through this device, we highlight the benefits of leads that can stimulate through smaller, directional contacts. Those benefits include that smaller electrodes increase the selectivity of smaller axons compared to larger 68 Figure 4.7. Smaller contact sizes more efficiently activate small-diameter fibers. A. Multicompartment axon models were run with diameters of 2.0 µm, 5,7 µm, and 10.0 µm in response to stimulation from 1 to 36 contacts on the µDBS. B. Firing threshold was normalized to -1 V amplitude at nine contacts turned on, which is approximately the height of classic DBS contact of 1.5 mm. A multiresolution device may be useful to target different diameter fibers; as shown in the right panel, smaller contacts activate smalldiameter fibers at 65.75% efficiency over 5.7 µm fibers, and require about 113.1% additional voltage to activate the same 10.0 µm fibers. Larger contact sizes preferentially activate large-diameter fibers, with about 50% lower thresholds relative to 5.7 µm fibers. 69 ones. This change in relative selectivity could be used to improve therapeutic DBS since smaller fibers are more associated with clinical benefit whereas larger fibers are more associated with side effects (Chaturvedi et al., 2010; Lang et al., 1999). The novel design of the µDBS does not merely enable smaller contacts, but also combine those contacts to accommodate larger electrodes than what is clinically available. That the total surface area depends on how many contacts are tied to the same bus line affords the µDBS an unprecedented level of flexibility, which could be useful in both research and clinical applications. In the validation tests we conducted on the device, the programming experiment in Figure 4.5 demonstrates that errors, which might arise and incorrectly assign contacts to bus lines, are device dependent. Some chips do not display any errors, regardless of a longer or shorter programming time. With our setup, the fastest programming time we achieved was ~2.7 s for 216 bits; however, there was no trend on programming error rate and programming time. This finding leads us to believe that programming can be achieved at faster frequencies and still program with perfect fidelity. With this in mind, programming the µDBS device would be feasible if it were to be applied in the clinic, and our technology provides a method that could reasonably be used to expand the number of contacts on clinical µDBS devices. However, given that it would be impossible to manually choose optimal contact configurations for such a device, we have previously published an optimization algorithm that can identify optimal contact amplitudes and configurations in near real-time (D. N. Anderson et al., 2018). Finally, the impedances reported for individual contacts were around 180 kΩ on average for three chips, but single contact impedances vary on a chip-by-chip basis, which 70 may be the result of slight variations in the postprocessing of chips. For high impedances, it would be possible to safely stimulate only with low voltages, based on FDA standards of charge density. However, the advantage of this technology is in multiresolution stimulation, in which contacts can be grouped together. Contacts grouped together function as larger electrodes and can mimic the size of clinical electrodes, having similar impedances to those recorded in the clinic. Because only flat chips were tested, the entire µDBS chip has the ability to quadruple the possible surface area and further reduce the impedance. The differing impedance levels recorded demonstrate that contacts can be recruited appropriately. The bath testing recorded voltage profiles with two groups of contacts tied to one of two active bus lines, with one bus line set to twice the amplitude of the other. When the bus lines were swapped, there was a notable shift in the voltage field, which demonstrates how the µDBS is able to recruit contacts to separate bus lines simultaneously. The chapter reports on validation tests on flat chips, of which four are needed for the full device. However, additional steps must be completed to package the complete device and ensure longevity in tissue. For the chips tested, silicone was used to encase wirebonds and traces exposed on the PCB in the saline bath, but better packaging must be done for use in chronic animal studies. Additionally, platinum is the preferred metal for tissue interfaces and may offer better longevity and stability than gold in a tissue environment. For the computational experiments, we limited our study to vertical neurons to match the orientation of the contacts on the µDBS we studied, but as we previously have shown, different neuron orientations can change the activation profiles (D.N. Anderson et al., 2019). 71 4.6 Conclusions The technology we developed in the µDBS is meant to push the field of DBS technology from the initial DBS design that has been used for decades now toward directional leads with a greater number of smaller contacts. The chapter lays the groundwork for the technology required to increase lead complexity, which would allow for more contacts without the addition of more wires. This chapter also proposed the concept of a multiresolution DBS device, with a range of large to small contacts to enable either widespread or more restricted, selective stimulation. Finally, we show novel evidence that smaller, directional contacts may be better for stimulation therapy than once thought: not only is there greater field shaping flexibility with directional contacts, but smaller contacts also improve targeting of smaller diameter fibers, which may lead to increases in the therapeutic window. 4.7 References Anderson, C. J., Anderson, D. N., Pulst, S. M., Butson, C. R., & Dorval, A. D. (2019). Deep brain stimulation dose equivalence and maximizing efficacy with pulse width tuning and directional electrodes. Manuscript submitted. Anderson, D. N., Duffley, G., Vorwerk, J., Dorval, A. D., & Butson, C. R. (2019). Anodic stimulation misunderstood: Preferential activation of fiber orientations with anodic waveforms in deep brain stimulation. Journal of Neural Engineering, 16(1), 016026. Anderson, D. N., Osting, B., Vorverk, J., Dorval, A. D., & Butson, C. R. (2018). Optimized programming algorithm for cylindrical and directional deep brain stimulation electrodes. Journal of Neural Engineering, 15(2), 026005. Butson, C. R., Maks, C. B., & McIntyre, C. C. (2006). Sources and effects of electrode impedance during deep brain stimulation. Clinical Neurophysiology, 117(2), 447– 454. https://doi.org/10.1016/j.clinph.2005.10.007 Chaturvedi, A., Butson, C. R., Lempka, S. F., Cooper, S. E., & McIntyre, C. C. (2010). 72 Patient-specific models of deep brain stimulation: Influence of field model complexity on neural activation predictions. Brain Stimulation, 3(2), 65–77. https://doi.org/10.1016/j.brs.2010.01.003 Contarino, M. F., Bour, L. J., Verhagen, R., Lourens, M. A. J., de Bie, R. M. A., van den Munckhof, P., & Schuurman, P. R. (2014). Directional steering: A novel approach to deep brain stimulation. Neurology, 83(13), 1163–1169. https://doi.org/ 10.1212/WNL.0000000000000823 Eisinger, R. S., Cernera, S., Gittis, A., Gunduz, A., & Okun, M. S. (2019). A review of basal ganglia circuits and physiology: Application to deep brain stimulation. Parkinsonism & Related Disorders. https://doi.org/10.1016/j.parkreldis. 2019.01.009 Lang, A. E., Lozano, A. M., Ashby, P., Kumar, R., & Kim, Y. J. (1999). Neurophysiological effects of stimulation through electrodes in the human subthalamic nucleus. Brain, 122(10), 1919–1931. https://doi.org/10.1093/ brain/122.10.1919 Lanotte, M. M., Rizzone, M., Bergamasco, B., Faccani, G., Melcarne, A., & Lopiano, L. (2002). Deep brain stimulation of the subthalamic nucleus: Anatomical, neurophysiological, and outcome correlations with the effects of stimulation. Journal of Neurology, Neurosurgery &amp; Psychiatry, 72(1), 53. https://doi.org/10.1136/jnnp.72.1.53 McIntyre, C. C., Richardson, A. G., & Grill, W. M. (2002). Modeling the excitability of mammalian nerve fibers: Influence of afterpotentials on the recovery cycle. Journal of Neurophysiology, 87(2), 995. Merrill, D. R., Bikson, M., & Jefferys, J. G. (2005). Electrical stimulation of excitable tissue: Design of efficacious and safe protocols. Journal of Neuroscience Methods, 141(2), 171–198. Pollo, C., Kaelin-Lang, A., Oertel, M. F., Stieglitz, L., Taub, E., Fuhr, P., … Schüpbach, M. (2014). Directional deep brain stimulation: An intraoperative double-blind pilot study. Brain, 137(7), 2015-2026. https://doi.org/10.1093/brain/awu102 Rebelo, P., Green, A., Aziz, T., Kent, A., Schafer, D., Venkatesan, L., & Cheeran, B. (2018). Thalamic directional deep brain stimulation for tremor: Spend less, get more. Brain Stimulation, 11(3), 600–606. Steigerwald, F., Müller, L., Johannes, S., Matthies, C., & Volkmann, J. (2016). Directional deep brain stimulation of the subthalamic nucleus: A pilot study using a novel neurostimulation device. Movement Disorders, 31(8), 1240–1243. https://doi.org/10.1002/mds.26669 73 Willsie, A. C., & Dorval, A. D. (2015a). Computational field shaping for deep brain stimulation with thousands of contacts in a novel electrode geometry. Neuromodulation: Technology at the Neural Interface, 18(7), 542–551. Willsie, A. C., & Dorval, A. D. (2015b). Fabrication and initial testing of the µDBS: A novel deep brain stimulation electrode with thousands of individually controllable contacts. Biomedical Microdevices, 17(3), 56. https://doi.org/10.1007/s10544-0159961-x CHAPTER 5 CONCLUSION Deep brain stimulation (DBS) shows great potential for treating many disorders outside the realm of Parkinson’s disease and essential tremor. As more disorders appear treatable through DBS, it is even more important that the field fully understands DBS programming. In many novel DBS applications, symptoms may be psychiatric or transient, making it nearly impossible to achieve optimal therapy based on acute timeframe symptom evaluation alone. However, the DBS programming advances demonstrated in this dissertation may make it possible to improve neural selectivity and help identify the specific targets responsible for therapeutic benefit in new neurological indications. Further, these increases in selectivity through DBS programming could possibly lead to improved rates of success in DBS for established disorders. In the dissertation, we have discussed numerous means by which DBS programming could be improved. First, we discuss the differential activation effects of cathodic versus anodic stimulation and identify a previously unknown relationship between fiber orientation and the anodic and cathodic contribution in the stimulus waveform. This improved understanding of the selective effects through anodic or cathodic stimulation may lead to improvements in tuning of the activation field through DBS programming, enabling better targeting of fibers of different orientations surrounding DBS electrodes. Using these 75 new principles found in Chapter 2, we have developed an optimization algorithm that can take into account fiber orientation in targeting neurons. Such an algorithm optimizes stimulation for neural targets identified by the user in near real-time by sifting through the parameter space of all contact configurations and amplitudes for a variety of lead designs. Finally, we propose and demonstrate the construction and testing of a novel DBS electrode, the µDBS, with microscale contacts that may enable selectivity toward smaller diameter, therapeutic fibers. Additionally, this device offers multiscale stimulation contacts that range in orders of magnitude in size to increase flexibility in DBS programming that may be useful in generating precise stimulation fields for established or novel DBS applications. 5.1 Anodic Versus Cathodic Stimulation Our findings from Chapter 2 show that cathodic and anodic stimulation have differing effects on neuron activation patterns based on the orientation of the axons surrounding the lead. Classically, anodic stimulation has been understood to be less likely to activate neurons and requires much larger thresholds — approximately five times higher — to achieve the same levels of activation in computational models (Rattay & Aberham, 1993). However, what we uncover through this research is that anodic and cathodic stimulation activates separate populations of neurons based on fiber orientation. Favorable orientations of anodic stimulation include axons that orthogonally approach or leave the electrode source, while segments of axons that pass tangential to the source are preferentially activated by cathodic stimulation. We find that anodic stimulation exclusively activates the orthogonal segments of neurons, and cathodic stimulation exclusively activates passing axon segments. Additionally, anodic stimulation activates 76 orthogonal axon segments at much lower thresholds than those at which cathodic stimulation activates passing segments. In principle, it may be possible to consider axon orientations when programming DBS devices. The active electrode could be programmed as an anode or cathode based on the fiber orientations surrounding the contact; alternately, the stimulation waveform — which has anodic and cathodic components — could be adjusted to either favor or reduce activation of one fiber orientation over another. This consideration could be especially advantageous, as we have shown in Chapter 2, if fiber tracts responsible for clinical benefits and side effects are in different orientations relative to the stimulation source. Additionally, this principle of orientation selectivity has implications for bipolar stimulation, which uses at least one contact as an anode and at least one contact as a cathode. There are instances in which patients do not receive therapeutic benefit from monopolar stimulation, but do experience therapy through bipolar stimulation. One hypothesis we propose for this phenomenon is that the addition of the anode contact in a bipolar configuration alters the fiber tracts that are being activated based on orientation and may select against fibers that cause side effects. It must be considered that neural selectivity based on fiber orientation adds an additional dimension in terms of DBS programming, further complicating programming, but this consideration could be worth taking into account for challenging programming cases. 5.2 Optimization of DBS Parameters The goals of the optimization algorithm proposed in this work were twofold: to reduce programming time/burden on programming physicians and to pave the way for 77 novel electrode designs with greater numbers of directional contacts. It has been reported that it takes tens of hours of programming time for DBS in the first year postimplant (Hunka, Suchowersky, Wood, Derwent, & Kiss, 2005; Ondo & Bronte-Stewart, 2005). Part of the reason for the lengthy duration of DBS programming is the trial-and-error nature of DBS programming, as well as the nuance in programming for nonsimple cases of DBS. In Parkinson’s disease, patients typically see a 50% improvement in motor scores (Benabid, Chabardes, Mitrofanis, & Pollak, 2009); however, patients do not experience a therapeutic effect in many instances and must undergo revision surgery. Approximately 15.2% - 34.0% of DBS procedures are for revision or removal of intracranial electrodes, and up to 48.5% of these revision surgeries may be due to lack of effect (Rolston, Englot, Starr, & Larson, 2016). Complicated cases of DBS could include challenging lead placements, where the lead has missed its target and ventured too close to a region associated with side effects. Burchiel et al. (2013) has reported that lead deviations of a few millimeters are common whereas deviations of greater than 3 mm from the target would result in suboptimal or ineffective DBS therapy (Ellis et al., 2008; Guridi et al., 2000). DBS programming is largely performed without visualization of the specific nuclei and fiber tracts surrounding the lead, but in such challenging DBS programming cases, it may be advantageous to visualize nearby structures using patient imaging to guide DBS targeting. We believe that patient-specific visualization of neural structures surrounding the electrode is the first step to improving neural targeting through DBS programming. At the same time, it remains a challenge to know which precise settings could optimally steer activation toward therapeutic structures and away from regions of avoidance. Therefore, 78 we created an optimization algorithm that automates — in near real-time — the selection of contacts and amplitudes based on known regions of therapy and regions of avoidance. Our algorithm is based in the second derivative of voltage along a neuron, also known as the activating function, to quickly approximate action potential generation (Rattay, 1986, 1999). We expanded the activating function to three dimensions using the Hessian matrix, which also allows us to apply fiber orientation selectivity from our findings from Chapter 2 to the optimization algorithm. The adoption of such an algorithm in DBS programming can accommodate the growing specificity of neural targets and the complexity of electrode designs as the DBS field advances and new targets are uncovered and directional leads come onto the market. The specificity and complexity of the targeting conditions, which cause a time burden for a programming clinician, are addressed by the algorithm in near real-time. At the very least, the optimization algorithm could provide the programming clinician with an educated starting point of DBS settings that could be tuned further if needed, reducing the burden on the programming clinician and patient, true to our objectives for this algorithm. Our second motivating factor for the optimization algorithm is to support the growing lead complexity in recent years. In the past few years, two new directional leads have entered the market, and even though they offer simple changes to the original quadripolar design, the number of total contacts doubles from four to eight contacts. This doubling of contacts increases the parameter space by orders of magnitude, and further increases in the number of contacts will make the programming of these devices infeasible. The algorithm we propose in this dissertation easily computes optimal solutions for these directional devices. Not only would applying this algorithm result in a reduction in 79 programming burden, but it could also bring about contact configurations that clinicians would not likely arrive at, based on the sheer number of possible parameters combinations. Until such automation is integrated into DBS programming, it is unlikely that there will be further advances in the DBS lead technology. Incorporation of such an algorithm would help remove at least one of the barriers to creating directional leads with greater numbers of smaller contacts. 5.3 Highly Directional Deep Brain Stimulation Our motivation to create the µDBS, a novel electrode with hundreds of microscale contacts, is to create a substantially more directional lead than those currently available. Even more importantly, the device we discuss features multiresolution contact sizes, in which the 864 contacts on the µDBS can work concurrently to create contacts functionally much larger than each individual one on its own. This device provides the DBS field with the first instance of a multiresolution DBS device that can stimulate tissue with contact sizes similar to those found in the clinic in classic cylindrical leads (6 mm2) or with novel directional leads (2 mm2) but that is also capable of generating more complex fields or narrow and unusual activation shapes through the use of much smaller contacts. Regardless of the neurological disease that is being treated, the shape and size of the contacts are flexible and can be optimized on a case-by-case basis. Creating a device like the µDBS lays the groundwork for possible improvements to expanding the total number of contacts a chip can have without being limited to the number of wires that fit within a lead shaft. Using such a device greatly increases the programming flexibility possible in DBS. At the same time, it would be impossible for anyone to sift through all 80 possible parameter combinations given the number of contacts on the µDBS. However, programming such a device becomes possible when combined with the optimization algorithm from Chapter 3. Additionally, because the optimization algorithm can assign different voltages to different contacts, the algorithm will perfectly translate to the µDBS to generate either multiple monopolar voltage states or complex bipolar configurations given the seven independent buslines on the µDBS. In Chapter 4, we also offer evidence through computational modeling that indicates that directional leads may have an additional benefit beyond the directional steering of activation fields. We found that using smaller, directional contacts may also increase the therapeutic window by more efficiently activating small-diameter fibers and avoiding large-diameter fibers. Even though we do not always know the precise neural targets — including the axon diameter in those neural targets — for many disorders that could be treated by DBS, there are already opportunities to realize benefits from using smaller contacts in established DBS applications. In Parkinson’s disease, for example, the most common treatment target is the subthalamic nucleus, which is reported to have neurons 12 µm in diameter in contrast to the 5.7-10 µm fibers of the internal capsule that are associated with numerous side effects (Chaturvedi, Butson, Lempka, Cooper, & McIntyre, 2010; Lang, Lozano, Ashby, Kumar, & Kim, 1999). It has also been reported that directional stimulation widens the therapeutic window by steering stimulation to therapeutic regions and away from side-effect-inducing regions (Contarino et al., 2014; Dembek et al., 2017; Pollo et al., 2014). However, we propose an additional mechanism: smaller directional contacts more efficiently activate the smaller diameter fibers in the STN rather than passing large fibers associated with side effects, which leads to increases in the 81 therapeutic window. Directional contacts used in the clinic could be further reduced in size to increase this selectivity effect for fiber diameter, and using a multiresolution DBS design like the µDBS could be the solution to enable targeting of smaller neurons yet maintain the ability to stimulate through contact sizes currently used in the clinic. 5.4 DBS Programming: Moving Forward 5.4.1 Using Anodic Stimulation in DBS Programming Some patients already implanted with DBS who demonstrate suboptimal outcomes may experience improvements in therapy through anodic stimulation, especially in challenging cases where monopolar cathodic stimulation or bipolar stimulation is not found to be effective. Based on recent studies in which anodic stimulation was experimentally used, some patients found greater therapeutic improvements with anodic stimulation rather than cathodic stimulation (Kirsch, Hassin-Baer, Matthies, Volkmann,& Steigerwald, 2018). At the moment, only leads from Boston Scientific are able to be programmed with monopolar anodic stimulation using a standard programmer. However, all devices currently on the market are technically capable of monopolar anodic stimulation, yet they are unable to do this in practice because of manufacturer restrictions. Our computational results, and outcomes from early human studies, indicate that monopolar anodic stimulation could be beneficial to some patients immediately. Since we have discovered that anodic stimulation targets fibers of different orientations compared to cathodic stimulation, it would be possible to use this principle of selectivity to avoid fibers that cause side effects based on fiber orientation with either anodic or cathodic stimulation. We would strongly encourage the DBS industry to incorporate anodic stimulation into the realm of 82 programming possibilities in DBS devices, as such an adjustment could benefit patients who do not experience ample improvement from monopolar cathodic settings or bipolar settings. Additionally, a capability for devices to use anodic stimulation could be retroactively applied to previously implanted devices. We would, further, encourage clinicians to consider incorporating monopolar anodic stimulation in their programming when possible given that there already exist patients who have exhibited greater therapeutic benefit from anodic stimulation than from cathodic stimulation. For now, however, we believe that more studies should be conducted to explore and characterize the effects of anodic stimulation on fibers surrounding stimulation contacts and how those activation profiles may lead to therapeutic outcomes different from those from cathodic stimulation. 5.4.2. DBS Programming for Targeting Networks DBS is highly successful in treating movement disorders like Parkinson’s disease and essential tremor, but it has been much more challenging to translate this success to other disorders. Despite preliminary evidence of improvement for a variety of other neurological disorders using DBS, improvements have been far more variable in Tourette’s and depression (Maciunas et al., 2007; Malone et al., 2009), for example. One explanation as to why DBS works well in movement disorders like Parkinson’s disease and essential tremor is that targets of those disorders are well known and established, and protocols for patient screening for DBS candidates for these disorders are well defined. In recent years, it has been hypothesized that disorders like Parkinson’s disease and essential tremor may have a common or highly conserved disease network across many patients that is simpler to target, and that successful targeting of the disease network is responsible for therapeutic 83 outcomes (Horn et al., 2017). This network targeting hypothesis may also explain why only ~10% of patients with disorders like epilepsy – which may have highly variable epileptic circuits – achieve substantial therapeutic benefit for at least one year following the onset of stimulation (Fisher et al., 2010; Jobst et al., 2017). In a similar vein, variable patientspecific networks may be responsible for the difficulty in establishing common targets for psychiatric disorders. For patients who undergo DBS therapy for treatment-resistant depression, the variability in patient responses may be due to different network irregularities that happen to cause similar depressive symptoms across patients. Not knowing the pathology of a patient’s depression makes identifying a target for stimulation very difficult. Such variability across patients with such disorders might make a universal strategy for targeting difficult to achieve without novel, patient-specific targeting approaches. One strategy to address issues in advancing DBS for other disease applications without specific nucleus or fiber tract targets is to use the optimization algorithm to target networks rather than neural structures. To target a disease network, the algorithm could be applied to target connectivity metrics correlated with clinical benefit (Horn et al., 2017; Gu et al., 2015; Muldoon et al., 2016). Targeting a network may be difficult to do manually given the complexity of neural networks and the added level of abstraction in targeting functional networks, but the optimization algorithm could be a reliable way to target disease networks. Additionally, an objective and reliable way to target and avoid networks could lead to improvements in target identification by potentially isolating the common factor in successful treatment through well-controlled clinical studies. Finally, programming patients for Parkinson’s disease and essential tremor is much 84 more straightforward than programming for psychiatric disorders, given the fast time course of symptom response and the fact that patients will generally show improvements immediately with effective settings. Trial-and-error programming becomes much more challenging for disorders with transient symptoms, such as in epilepsy, where patients are not actively having seizures during the programming session. For psychiatric disorders, such as treatment-resistant depression or obsessive compulsive disorder, response times may take months to manifest, if they do at all in a given setting. Situations in which a clinician cannot rely on feedback or visual cues during a programming session motivate a new tool, such as our optimization algorithm, to reliably stimulate the neural targets theorized to have therapeutic benefit. 5.4.3. Pushing DBS Programming and DBS Technology Forward In this dissertation, we took three approaches to guiding selective neural targeting through DBS programming: determining specific targeting of fiber orientations through changes in waveform, designing an algorithm to optimize targeting of nuclei and fiber tracts, and creating a novel, multiresolution DBS device that can more specifically focus the stimulation effect on small-diameter fibers. The µDBS is unlike other DBS electrodes previously commercialized or published on: it uniquely follows the design of a microgrid array applied in the context of a DBS electrode, which has classically used millimeter scale contacts. For decades, DBS lead technology has been largely unchanged, but now an abundance of evidence indicates that directional leads offer benefits over classic, cylindrical contact designs. With directional leads, one could steer stimulation away from side-effect-inducing regions to regions responsible for therapeutic benefit. Properly 85 utilized directional steering would make it difficult to induce side effects and effectively widen the window of therapy. On the µDBS, we increased the number of contacts by orders of magnitude past what is available to enable very fine control of stimulation fields. Through the technology we presented in Chapter 4, clinical DBS leads will no longer be limited in the number of contacts they could have, since each contact no longer requires its own wire. The most important feature of our directional DBS device is the novel use of multiscale contacts in DBS, which has thus far been restricted to contacts of fixed size. This multiresolution concept could also be applied to neural recordings, with large contacts used for measuring local field potentials or single contacts to capture single unit recordings. Further work is left to make the µDBS ready for in vivo studies. In advance of in vivo studies, it may be possible to test the µDBS functionality in slice or an in vitro nerve preparation. The µDBS can have many applications that are not restricted to the central nervous system. For example, single, flat chips of the µDBS may have an application in the periphery for the targeting of nerves. In all instances of testing, however, the assembly and packaging of the device is key for its use in a harsh, tissue environment. We hope to use the computational work in Chapter 4 to encourage DBS manufacturers to make DBS electrodes with a greater number of smaller contacts, given that we have shown that smaller diameter fibers are more efficiently activated through smaller contacts. Additionally, the grid-like arrangement of contacts on the µDBS could lead to fiber selectivity between vertical versus tangential fibers based on whether a row or a column of contacts is activated. This new principle of neural selectivity based on orientation takes what we have found in Chapter 2 even further, since vertical and tangential neurons are equally activated by cathodic stimulation. Using electrodes in either 86 row or column arrangements could favor vertical versus tangential fibers, which could further be used for neural selectivity in DBS programming. Considering all aspects of selective neural targeting through DBS programming discussed in this dissertation, the parameter space is simply too great for a clinician to simultaneously address, especially with the nearly infinite parameter combinations on the µDBS. In the future, our next step will be to combine the µDBS with the optimization algorithm so that all of the nuances of neural targeting can be taken into account during DBS programming without an additional burden for the programming physician. 5.5 Final Conclusions We have demonstrated that neurons can be preferentially activated based on fiber orientation with simple changes to DBS programming; we have specifically found that cathodic stimulation targets passing fibers and anodic stimulation targets fibers approaching or leaving the electrode. Considering the role of fiber orientation on activation, we present an optimization algorithm to automate the selection of patientspecific DBS parameters for selective neural targeting in near real-time depending on userdefined targets of interest and avoidance. Finally, we worked to further the field of DBS lead technology by enabling finer control over stimulation profiles and precise targeting or avoidance of neural regions during DBS programming through the creation of the µDBS, a novel directional DBS lead with hundreds of microscale contacts. The µDBS has contacts orders of magnitude smaller than contacts of modern DBS leads, and we demonstrate that smaller contacts selectively activate smaller diameter fibers, which may lead to increases in the window of therapy. Taken together, this body of work represents a major 87 advancement toward optimizing DBS therapy and guiding neural selectivity through DBS programming, and we believe the work in this dissertation is likely to improve patient care long term. 5.6 References Benabid, A. L., Chabardes, S., Mitrofanis, J., & Pollak, P. (2009). Deep brain stimulation of the subthalamic nucleus for the treatment of Parkinson’s disease. The Lancet Neurology, 8(1), 67–81. https://doi.org/10.1016/S1474-4422(08)70291-6 Burchiel, K. J., McCartney, S., Lee, A., & Raslan, A. M. (2013). Accuracy of deep brain stimulation electrode placement using intraoperative computed tomography without microelectrode recording. Journal of Neurosurgery, 119(2), 301–306. https://doi.org/10.3171/2013.4.JNS122324 Chaturvedi, A., Butson, C. R., Lempka, S. F., Cooper, S. E., & McIntyre, C. C. (2010). Patient-specific models of deep brain stimulation: Influence of field model complexity on neural activation predictions. Brain Stimulation, 3(2), 65–77. https://doi.org/10.1016/j.brs.2010.01.003 Contarino, M. F., Bour, L. J., Verhagen, R., Lourens, M. A. J., de Bie, R. M. A., van den Munckhof, P., & Schuurman, P. R. (2014). Directional steering: A novel approach to deep brain stimulation. Neurology, 83(13), 1163–1169. https://doi.org/10.1212/WNL.0000000000000823 Dembek, T. A., Reker, P., Visser-Vandewalle, V., Wirths, J., Treuer, H., Klehr, M., … Timmermann, L. (2017). Directional DBS increases side-effect thresholds—A prospective, double-blind trial. Movement Disorders, 32(10), 1380–1388. https://doi.org/10.1002/mds.27093 Ellis, T.-M., Foote, K. D., Fernandez, H. H., Sudhyadhom, A., Rodriguez, R. L., Zeilman, P., … Okun, M. S. (2008). Reoperation for suboptimal outcomes after deep brain stimulation surgery. Neurosurgery, 63(4). Retrieved from http://journals.lww.com/neurosurgery/Fulltext/2008/10000/REOPERATION_FO R_SUBOPTIMAL_OUTCOMES_AFTER_DEEP.26.aspx Fisher, R., Salanova, V., Witt, T., Worth, R., Henry, T., Gross, R., … Labar, D. (2010). Electrical stimulation of the anterior nucleus of thalamus for treatment of refractory epilepsy. Epilepsia, 51(5), 899–908. Guridi, J., Rodriguez-Oroz, M., Lozano, A., Moro, E., Albanese, A., Nuttin, B., … Obeso, J. (2000). Targeting the basal ganglia for deep brain stimulation in Parkinson’s 88 disease. Neurology, 55(12 Suppl 6), S21–S28. Retrieved from PubMed. (11188971) Gu, S., Pasqualetti, F., Cieslak, M., Telesford, Q. K., Yu, A. B., Kahn, A. E., … Bassett, D. S. (2015). Controllability of structural brain networks. Nature Communications, 6, 8414. Horn, A., Reich, M., Vorwerk, J., Li, N., Wenzel, G., Fang, Q., … Fox, M. D. (2017). Connectivity predicts deep brain stimulation outcome in Parkinson disease. Annals of Neurology, 82(1), 67–78. https://doi.org/10.1002/ana.24974 Hunka, K., Suchowersky, O., Wood, S., Derwent, L., & Kiss, Z. H. T. (2005). Nursing time to program and assess deep brain stimulators in movement disorder patients. Journal of Neuroscience Nursing, 37(4). Retrieved from http://journals.lww.com/jnnonline/Fulltext/2005/08000/Nursing_Time_to_Progra m_and_Assess_Deep_Brain.6.aspx Jobst, B. C., Kapur, R., Barkley, G. L., Bazil, C. W., Berg, M. J., Bergey, G. K., … Duchowny, M. S. (2017). Brain-responsive neurostimulation in patients with medically intractable seizures arising from eloquent and other neocortical areas. Epilepsia, 58(6), 1005–1014. Kirsch, A. D., Hassin-Baer, S., Matthies, C., Volkmann, J., & Steigerwald, F. (2018). Anodic versus cathodic neurostimulation of the subthalamic nucleus: A randomized-controlled study of acute clinical effects. Parkinsonism & Related Disorders. https://doi.org/10.1016/j.parkreldis.2018.05.015 Lang, A. E., Lozano, A. M., Ashby, P., Kumar, R., & Kim, Y. J. (1999). Neurophysiological effects of stimulation through electrodes in the human subthalamic nucleus. Brain, 122(10), 1919–1931. https://doi.org/10.1093/brain/122.10.1919 Maciunas, R. J., Maddux, B. N., Riley, D. E., Whitney, C. M., Schoenberg, M. R., Ogrocki, P. J., … Gould, D. J. (2007). Prospective randomized double-blind trial of bilateral thalamic deep brain stimulation in adults with Tourette syndrome. Journal of Neurosurgery JNS, 107(5), 1004–1014. https://doi.org/10.3171/JNS-07/11/1004 Malone, D. A., Dougherty, D. D., Rezai, A. R., Carpenter, L. L., Friehs, G. M., Eskandar, E. N., … Greenberg, B. D. (2009). Deep brain stimulation of the ventral capsule/ventral striatum for treatment-resistant depression. Biological Psychiatry, 65(4), 267–275. https://doi.org/10.1016/j.biopsych.2008.08.029 Muldoon, S. F., Pasqualetti, F., Gu, S., Cieslak, M., Grafton, S. T., Vettel, J. M., & Bassett, D. S. (2016). Stimulation-based control of dynamic brain networks. PLoS Computational Biology, 12(9), e1005076. 89 Ondo, W. G., & Bronte-Stewart, H. (2005). The North American Survey of placement and adjustment strategies for deep brain stimulation. Stereotactic and Functional Neurosurgery, 83(4), 142–147. Pollo, C., Kaelin-Lang, A., Oertel, M. F., Stieglitz, L., Taub, E., Fuhr, P., … Schüpbach, M. (2014). Directional deep brain stimulation: An intraoperative double-blind pilot study. Brain, 137(7), 2015-2026. https://doi.org/10.1093/brain/awu102 Rattay, F. (1986). Analysis of models for external stimulation of axons. IEEE Transactions on Biomedical Engineering, BME-33(10), 974–977. https://doi.org/10.1109/ TBME.1986.325670 Rattay, F. (1999). The basic mechanism for the electrical stimulation of the nervous system. Neuroscience, 89(2), 335–346. https://doi.org/10.1016/S0306-4522(98)00330-3 Rattay, F., & Aberham, M. (1993). Modeling axon membranes for functional electrical stimulation. IEEE Transactions on Biomedical Engineering, 40(12), 1201–1209. https://doi.org/10.1109/10.250575 Rolston, J. D., Englot, D. J., Starr, P. A., & Larson, P. S. (2016). An unexpectedly high rate of revisions and removals in deep brain stimulation surgery: Analysis of multiple databases. Parkinsonism & Related Disorders, 33, 72–77. https://doi.org/10.1016/j.parkreldis.2016.09.014