Real-time suppression and amplification of frequency-specific neural activity using stimulation evoked oscillations

Abstract

Background: Approaches to predictably control neural oscillations are needed to understand their causal role in brain function in healthy or diseased states and to advance the development of neuromodulation therapies.Objective: We present a closed-loop neural control and optimization framework to actively suppress or amplify low-frequency neural oscillations observed in local field potentials in real-time by using electrical stimulation. The rationale behind this control approach and our working hypothesis is that neural oscillatory activity evoked by electrical pulses can suppress or amplify spontaneous oscillations via destructive or constructive interference when the pulses are continuously delivered with appropriate amplitudes and at precise phases of the modulated oscillations in a closed-loop scheme. Methods: We tested our hypothesis in two nonhuman primates that exhibited a robust increase in lowfrequency (8e30 Hz) oscillatory power in the subthalamic nucleus (STN) following administration of the neurotoxin 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP). To test our neural control approach, we targeted 8e17 Hz oscillations and used electrode arrays and electrical stimulation waveforms similar to those used in humans chronically implanted with brain stimulation systems. Stimulation parameters that maximize the suppression or amplification of neural oscillations were predicted using mathematical models of the stimulation evoked oscillations. Results: Our neural control and optimization approach was capable of actively and robustly suppressing or amplifying oscillations in the targeted frequency band (8e17 Hz) in real-time in the studied subjects. Conclusions: The results from this study support our hypothesis and suggest that the proposed neural control framework allows one to characterize in controlled experiments the functional role of frequencyspecific neural oscillations by using electrodes and stimulation waveforms currently being employed in humans.

Deep brain stimulation pulse sequences to optimally modulate frequency-specific neural activity

“…We showed that when the amplitude envelope of the spontaneous oscillations is constant over time, phase modulation (or locking) is sufficient to achieve optimal suppression or amplification of the targeted oscillations. However, the envelope of real in-vivo spontaneous oscillations is not constant over time (1–4, 11, 12). Our results show that adjustments in the stimulation pulse frequency and amplitude are needed to create precise destructive or constructive interference between stimulation-evoked responses and spontaneous oscillations whose amplitude envelope varies over time.…”

“…The input x ( k ) = − 1 corresponds to stimulation pulse with the largest allowed amplitude and x ( k ) = 0 corresponds to the stimulation current that does not evoke a neural response. The values of the stimulation variable are negative or zero because the negative phase of the stimulation pulse (cathodal stimulation) was assumed to evoke the neural response (11). It is noteworthy that using the same modeling framework but with a change in the saturation limits one could also characterize evoked responses dominated by the anodal phase of the stimulus.…”

“…1. This model structure has been 103 shown to accurately approximate the linear response of neural 104 circuitry to electrical stimulation pulses (11,12). The evoked 105 response is described by the following equations of differences.…”

Toolkit for Oscillatory Real-time Tracking and Estimation (TORTE)

“…Others are accurate but TORTE require special hardware that is not easily maintained without dedicated engineering staff. They may be prohibited by high cost and may be difficult to implement into existing experimental protocols (Kanta et al, 2019;Rodriguez Rivero and Ditterich, 2021;Escobar Sanabria et al, 2020;Zrenner et al, 2018;Shirinpour et al, 2020). Many solutions are built atop proprietary, closed-source software such as MAT-LAB and its toolboxes (Hassan et al, 2020;Zelmann et al, 2020).…”

Stable modulation of neuronal oscillations produced through brain-machine equilibrium

“…and A.S. patent application WO/2020/183152), which we implemented as a digital circuit using the existing hardware of a commercially available recording system. Ordinarily, to produce such an online phase estimate the signal would be filtered with a pass band filter before a phase estimation step such as the Hilbert transform or the detection of zero crossings (Brittain et al, 2013;Cagnan et al, 2017;Escobar Sanabria et al, 2020;Kanta et al, 2019;Peles et al, 2020;Schreglmann et al, 2021;Siegle and Wilson, 2014;Takeuchi et al, 2021;Widge et al, 2018;Zanos et al, 2018). Such filters exhibit a filter delay, whereby the estimate at a given time point pertains to a fixed time in the past.…”