Proceedings of the 11th Annual Deep Brain Stimulation Think Tank: pushing the forefront of neuromodulation with functional network mapping, biomarkers for adaptive DBS, bioethical dilemmas, AI-guided neuromodulation, and translational advancements

Abstract

The Deep Brain Stimulation (DBS) Think Tank XI was held on August 9-11, 2023 in Gainesville, Florida with the theme of "Pushing the Forefront of Neuromodulation". The keynote speaker was Dr. Nico Dosenbach from Washington University in St. Louis, Missouri. He presented his research recently published in Nature inn a collaboration with Dr. Evan Gordon to identify and characterize the somato-cognitive action network (SCAN), which has redefined the motor homunculus and has led to new hypotheses about the integrative networks underpinning therapeutic DBS. The DBS Think Tank was founded in 2012 and provides an open platform where clinicians, engineers, and researchers (from industry and academia) can freely discuss current and emerging DBS technologies, as well as logistical and ethical issues facing the field. The group estimated that globally more than 263,000 DBS devices have been implanted for neurological and neuropsychiatric disorders. This year's meeting was focused on advances in the following areas: cutting-edge translational neuromodulation, cutting-edge physiology, advances in neuromodulation from Europe and Asia, neuroethical dilemmas, artificial intelligence and computational modeling, time scales in DBS for mood disorders, and advances in future neuromodulation devices.

Keywords: Parkinson's disease; adaptive DBS; artificial intelligence; deep brain stimulation (DBS); epilepsy; interventional psychiatry; neuroethics; optogenetics.

Figure 1

Somato-cognitive action network (SCAN). (A) Resting state functional connectivity (RSFC) seeded from the middle inter-effector node in primary motor cortex (bilaterally) in a representative individual (P1; 356 min resting-state fMRI). In cortex the SCAN includes three inter-effector nodes (superior, middle, inferior) that alternate with effector-specific foot, hand and mouth primary motor regions, as well as two nodes on the dorsal midline in the SMA (supplementary motor area) and dACC (dorsal anterior cingulate cortex) that are interleaved with the effector-specific regions of the SMA/pre-SMA. In thalamus, the centromedian nucleus [CM; black outline (Ewert et al., 2017)] is part of the SCAN. In the striatum, the dorsal posterior putamen forms part of the SCAN. In the cerebellum, crus VI and para-vermian VIIIA are part of the SCAN. Functional connectivity [Z(r)] in cortex was thresholded from 0.35 to 0.6, and from 0.15 to 0.3 to account for the lower signal-to-noise ratio. (B) In the integrate–isolate model of M1 organization, effector-specific—foot (green), hand (cyan) and mouth (orange)—functional zones are represented by concentric rings with proximal body parts surrounding the relatively more isolatable distal ones (toes, fingers and tongue). The SCAN inter-effector regions (maroon) sit at the intersecting points of these fields and support integrative, allostatic whole-body control.

Figure 2

Closed-loop vagus nerve stimulation (VNS) for reinforcement learning. VNS paired with successful behavioral outcome drives phasic cholinergic signaling from the basal forebrain and modulates neuronal representation of outcome within motor cortex. In addition, closed-loop VNS influences myelin plasticity and repair within motor cortex, restoring motor function in models of demyelination. Together, these results suggest a model in which closed-loop VNS can enhance behavioral reinforcement cues through cholinergic neuronal activity to promote circuit-specific plasticity and enhanced learning.

Figure 3

Example of adaptive DBS system for freezing of gait in Parkinson's disease using kinematic inputs. (A) Schematic of distributed aDBS system. Inertial measurement unit (IMU) data is streamed in real-time to a PC-in-the-loop. Gait arrhythmicity is measured from the IMU data. A decision to increase or decrease stimulation intensity is made depending on whether the measured arrhythmicity is above or below an established threshold, which is then communicated back to the Summit RC+S neurostimulator via the Summit communicator. (B) Measurement of the real-time arrhythmicity over the course of a stepping-in-place task relative to the threshold. (C) Stimulation decisions to increase or decrease stimulation based on the arrhythmicity. (D) Adaption of stimulation amplitude of the two subthalamic nuclei in response to the gait arrhythmicity. Adapted from Melbourne et al. (2023).

Figure 4

Differences in long term walking speed trends with and without DBS treatment. Gait dysfunction in PD is associated with the cholinergic system, and may be a more sensitive biomarker than other motor signs of PD that respond to medication or DBS. Plots show walking speed over a period of 2.5 years from 4 patients. Walking speed (meters/second) was captured using iPhone actigraphy during real world activities. Scatter points represent median walking speed during a 3 month window, shaded error bars represent the 25th and 75th percentile of walking speed respectively. Dotted line represents the date in which DBS was implanted. Data from 32,550 walking events, across 321.17 walking hours.

Figure 5

Subthalamic stimulation effects on emotional processing and alpha frequency. Intracranial subthalamic physiology recordings were paired with (A) a task-based emotional processing task with negative, positive and neutral images and recorded subjective valence and arousal (here shown with 1 second stimulation) (1). The results demonstrated the (B, C) expected late alpha desynchronization with affective imagery (1) which has been previously correlated with subjective valence and depression scores. (D) Using acute 1 second stimulation, there was a shift toward positive subjective valence bias to 10 Hz (alpha) stimulation but not 130 Hz (clinical) stimulation frequency (1) with (E) 10 Hz stimulation increasing alpha synchronization (i.e., loss of alpha desynchronization) (2). Both (F) acute 1 second 10 Hz stimulation and (G) subacute 15 min 10 Hz and 130 Hz stimulation of ventral contacts (3) was associated with greater positive emotional valence bias.

Figure 6

Mapping the effects of DBS for Tourette syndrome. (A) VTAs related to therapeutic stimulation and side effects. Each VTA color represents areas associated with the following effects: blue = therapeutic effect, orange = dizziness, green = paresthesia, and purple = depressed mood. The peach-colored region in the right hemisphere is the CM nucleus. (B) Normative connectome from VTAs related to therapeutic stimulation and side effects. The normative connectome from VTAs related to the therapeutic stimulation (above), and depressed mood (below). The peach-colored region is the CM nucleus. (C) Precision mapping of implanted deep brain stimulation electrodes in an atlas associated with tic improvements in the thalamus. The color bar represents the percentage improvement in tic symptoms.

Figure 7

The key stakeholders involved in and potential pathways by which participants in experimental neural device trials may lose access to their neural implant after the conclusion of a clinical trial.

Figure 8

Framework for bidirectional, surgeon and patient communication.

Figure 9

Fully automated, video-based motor assessment in Parkinson's disease. Upper half: (A–E) Stages of processing a patient video passes through to arrive at a disease severity rating. Lower half: The right panel shows an example video once processed. The upper left panel shows the output of activity recognition (right toe-tapping, followed by left toe-tapping, with no activity indicated in intervening periods). The white boxes show the clinical label (CUPDRS) and the model output (KUPDRS). The lower left panel shows the clinical signal extracted during a relevant activity (in this case, toe-tapping), the features of which are passed to a model for inference.

Figure 10

DBS fiber-filtering. Based on an anatomical model [pathway atlas or normative connectome, the basal ganglia pathway atlas (Petersen et al., 2019) is shown as an example, top panels], each streamline is selected and set into relationship with DBS stimulation volumes from a cohort of patients in a mass-univariate statistical approach. For each streamline, three general scenarios may result: (A) The streamline was predominantly modulated in patients in which a symptom of question was reduced, but less in patients in which the symptom deteriorated or stayed stable. If this is the case, a positive score is assigned to the streamline (these streamlines have been visualized with red color in most publications). (B) The opposite case: The streamline was predominantly modulated in the patients in which a symptom got worse (the tract receives a negative score and is often shown in blue). (C) There is no clear relationship between modulation of the streamline and changes of the symptom of question. In this case, the streamline is filtered out. When iteratively applied across all streamlines in the atlas, the bundles coding for improvements of a specific symptom can be identified (here the cerebellothalamic pathway for tremor improvement in right-lateralized DBS electrodes).