Using a virtual cortical module implementing a neural field model to modulate brain rhythms in Parkinson's disease

Abstract

We propose a new method for selective modulation of cortical rhythms based on neural field theory, in which the activity of a cortical area is extensively monitored using a two-dimensional microelectrode array. The example of Parkinson's disease illustrates the proposed method, in which a neural field model is assumed to accurately describe experimentally recorded activity. In addition, we propose a new closed-loop stimulation signal that is both space- and time- dependent. This method is especially designed to specifically modulate a targeted brain rhythm, without interfering with other rhythms. A new class of neuroprosthetic devices is also proposed, in which the multielectrode array is seen as an artificial neural network interacting with biological tissue. Such a bio-inspired approach may provide a solution to optimize interactions between the stimulation device and the cortex aiming to attenuate or augment specific cortical rhythms. The next step will be to validate this new approach experimentally in patients with Parkinson's disease.

Keywords: Parkinson's disease; brain stimulation; dysrhythmia; neural field model; neuroprosthetic devices.

Figure 1

Schematic representation of the cortical target and the MEA. The MEA provides a spatial mapping of neural activity in the cortical target compatible with the spatial scale of description of the neural field model (mesoscopic scale). Consequently, it is possible to exploit the neural field theory to process multi-site recordings provided by the MEA, and to combine these recordings into multiple stimulation signals aiming to modulate neuronal activity.

Figure 2

Localization, frequency analysis and coherence of primary motor cortex (M1) activity with tremor. Magnetoencephalographic (MEG) and electromyographic (EMG) data acquired in a PD patient with right hand tremor. The arrow shows where the MEG signal is recorded. The power spectrum of M1 MEG recording presents a strong peak centered around 10 Hz (middle). Cerebro-muscular (CM) coherence between M1 activity as measured by MEG and EMG of the Extensor Digitorum Communis (EDC) has a maximum around 10 Hz (Fig. modified from Timmermann et al., , with permission).

Figure 3

Principle of the proposed closed-loop stimulation method. Local Field Potentials (LFPs) are recorded by micro-electrodes of the high-density (on the order of 25 electrodes per square millimeter, similar to the Utah Array) MEA. These signals are independently amplified, filtered and analyzed to detect the presence of undesirable frequency components above a predetermined threshold. Stimulation is provided to the cortical target only at the spatial position(s) where the threshold is crossed.

Figure 4

Network activity (2D mapped to 1D) before (“control off”) and during (“control on”) closed-loop stimulation. “Pathological” 10 Hz oscillations of the recorded membrane potential (0–500 ms) are durably attenuated by the closed-loop stimulation signal (switched “on” between 500 and 1000 ms) controlling network activity in space and time. When the stimulation signal is present, the network is still active in other (“physiological”) frequency bands, i.e., network activity are not completely suppressed by the stimulation signal.

Figure 5

Power spectrum of neural activity before (blue) and during (red) stimulation. Before closed-loop stimulation, the dominant frequency is 10 Hz. During closed-loop stimulation specifically targeting the 10 Hz frequency, the power spectrum is strongly decreased at 10 Hz (and also at its sub-harmonics) whereas other frequency bands appear minimally affected. Note: (1) = strong attenuation of the 10 Hz peak, (2) and (3) = slight perturbation of the 50 and 85 Hz peaks respectively. This means that the closed-loop stimulation signal minimally interferes with the processing of inputs to the network at non-“pathological” frequencies.

Figure 6

Block diagram of the proposed virtual cortical module. The MEA is positioned on the cortex, and communicates via an interface INT with the chipset CPU. Recorded signals are amplified by the AMPL module, and then analyzed to detect the presence of “pathological” rhythms. If the undesirable rhythm is present, the TRIG module triggers the solving of neural field equations with the NFE module. The computed potential values are stocked in the BUFF module, to be later recalled when delayed values of the potential are needed. The potential values are then converted by the module CONV into a firing rate, i.e., a stimulation frequency. This stimulation frequency is transmitted to the STIM module that will trigger the stimulation to the cortical target. An internal clock SYNC synchronizes the various modules (Fig. from Beuter and Modolo, patent EP 09305432.8, 2009).