DBP: SIMULATION OF DEEP BRAIN STIMULATION Michael Okun, MD (Neurologist) Kelly Foote, MD (Neurosurgeon) Co-directors Center for Movement Disorders & Neurorestoration University of Florida Introduction and description/accomplishments of UFL DBS Center

DBS Overview Conventional 4 contact DBS Leads Brief introduction to DBS: over 100k patients implanted, and over 150 open trials on clinicaltrials.gov. After initial success and FDA approval for treating PD and ET, a wide range of other indications are being considered. The biggest public health problem that is being evaluated for DBS is depression, but it is also being tested for epilepsy, Alzheimer’s disease, obsessive-compulsive disorder, dystonia and others. Conventional 4 contact DBS Leads Wired Magazine, Issue 15.03, March 2007

Scientific Goals Accurate, patient-specific, multi-scale bioelectric field models of neuromodulation therapy. Deployment of mobile platform for DBS decision support. Development of virtual DBS surgery platform. Primary Tools

Recent Progress 2016 University of Utah Neuroscience Initiative awards a pilot grant to perform pre-operative imaging on Human Connectome PRISMA MRI scanners. Andrea Brock, 4th year Neurosurgery resident, is using this data to identify circuits modulated during effective versus ineffective stimulation.

Recent Progress 2016-2017 Collaboration with Dr. Foote on 17T imaging of human basal ganglia in DBS patient Bioengineering graduate student Katie Warthen is integrating imaging, field models and clinical DBS settings A P (V)

Recent Progress 2016-2017 International Neuromodulation Registry setup and received IRB approval in 2016. BRAIN Initiative R24 submitted Jan 2017 to develop tools and informatics platform to support BRAIN studies Internal advisory board includes faculty from SCI and Bioinformatics. External advisory board:

Challenges & Future Directions Near real-time, multi-scale bioelectric field modeling. Integration of models with outcomes and imaging databases. Growth of Infrastructure and methods to support population health studies in neuromodulation.

Clinical and Research Application of Transcranial Brain Stimulation Shorten into and motivate it more with a clinical scenario, sudden death and ischemia. Get to slide 4 in minute 3. Alvaro Pascual-Leone, MD, PhD, Michael Fox, MD, PhD, Mark Halko, PhD Beth Israel Deaconess Medical Center and Harvard Medical School

Motivation to Stimulate Cerebellum Nearly all known networks of the brain have cerebellar nodes Modified from Buckner et al 2011

TMS and Cerebellum The cerebellum is known for motor function, yet TMS is more likely to impact non motor function (orange: hypothetical TMS coil position, blue/red/green motor areas) Modified from Buckner et al 2011

TMS Position and Stimulation Positions of model TMS coil locations that could impact cerebellum Figure 2. Using finite element modeling, combined with resting state imaging, it is possible to localize the best location to position a coil to stimulate a cerebellar network node. Left: grid of potential stimulation sites tested in silico. Right: intensity of magnetic field reaching default network nodes from each site. Stimulating just cerebellar nodes was observed to change connectivity between two other default network nodes that were not directly stimulated themselves. Our design relies upon targeting resting-state networks for stimulation, an advance over traditional scalp location based targeting, and we leverage the distributed cortex-cerebellum connectivity to modulate entire networks [12]. Intensity of magnetic field reaching default network nodes from each candidate TMS site

Changing Network Functional Connectivity Stimulation of ideal default network site changes default network functional connectivity Default connectivity Changes in connectivity Modified from Halko et al 2014

Model predicts functional connectivity A single subject model field was used to assess the magnitude of effect upon cerebellar network nodes for individual subjects (e.g. each individual subject has slightly different locations of their cerebellar network “nodes”, thus recieving slightly different magnitudes of magnetic field). This is contrasted with the magnitude of functional connectivity change observed after stimulation.

Michael Fox, MD PhD: Electrodes and Targets Assistant Professor in Neurology 1) Two different electrode configurations used in the optimization. We note that head meshes in both configurations are identical except where the electrodes touch the scalp. 2) Cortical ROI derived from rsfMRI with SG as the seed point.

Results of Optimization Current E Field Optimization results for 2 methods on 2 electrode configurations. First row panels show the optimized stimulus pattern of a given method on a given electrode configuration. Second row shows the electric field normal to the cortex. Third row shows ERNI on the cortex. Error

Comparison of Approaches Electrode Current Optimization results for all four ROIs using two optimization criteria and two electrode configu- rations. Four cortical ROIs shown on the first column, with color representing positive (red) and negative (blue) correlation with four seed points: global pallidus pars internus (GPi), subthalamic nucleus (STN), subgenual cingulate (SG), and ventral intermediate nucleus (VIM). Each of the remaining panels shows both optimized electrode stimulus pattern and corresponding electric field component normal to the cortical surface. For better visibility, the electrodes with small currents in the optimal solution are not shown. Our results suggest that using a larger array of smaller electrodes does not necessarily yield more targeted modulation. For example, when Ruffini et al.’s least squares optimization criterion is used (i.e. the goal is to minimize the ERNI), optimized 27 electrode stimulus patterns yielded on average 35% smaller ERNI than optimized 82 electrode stimulus patterns. On the contrary, when our optimization criterion is used (i.e. the goal is to maximize directional current density in the ROI), 82 electrodes yielded on average 85% higher directional current in the ROI than 27 electrodes. These results suggest that different optimization criteria might favor different electrode number, type, and locations. Normal Cortical Field