Electrochemical and electrophysiological considerations for clinical high channel count neural interfaces

Abstract

Electrophysiological recording and stimulation are the gold standard for functional mapping during surgical and therapeutic interventions as well as capturing cellular activity in the intact human brain. A critical component probing human brain activity is the interface material at the electrode contact that electrochemically transduces brain signals to and from free charge carriers in the measurement system. Here, we summarize state-of-the-art electrode array systems in the context of translation for use in recording and stimulating human brain activity. We leverage parametric studies with multiple electrode materials to shed light on the varied levels of suitability to enable high signal-to-noise electrophysiological recordings as well as safe electrophysiological stimulation delivery. We discuss the effects of electrode scaling for recording and stimulation in pursuit of high spatial resolution, channel count electrode interfaces, delineating the electrode-tissue circuit components that dictate the electrode performance. Finally, we summarize recent efforts in the connectorization and packaging for high channel count electrode arrays and provide a brief account of efforts toward wireless neuronal monitoring systems.

Figure 1.

Comparisons between different types of electrodes and the Resolution-Coverage tradeoff. (a) Representative positioning of the different electrode types (surface electrocorticography [ECoG], depth, and penetrating surface) on the surface of the brain (illustrations not to scale). (b) Comparison of the inter-contact pitch, total coverage, and channel count offered by the state-of-the-art recording electrodes: 1-PtNRGrids, 2-Utah Array, 3-Neuropixels, 4-Neural Matrix, 5-Paradromics Argo,^, 6-NeuroGrid, 7-Viventi, 8-Escabi, 9-Ledochowitsch, 10-Molina-Luna, 11-Hollenberg, 12-Rubehn, 13-Kaiju, 14-Matsuo, 15-Toda, 16-Castagnola, 17-Zhao, 18- Precision Neuroscience, 19-Ad-Tech Medical Clinical Grid, 20-PMT Corporation Clinical Grid. The dashed region shows the tradeoff between the channel pitch and coverage for devices with limited channel count.

Figure 2.

Electrode material differences in impedance impact how the neural signal is recorded. (a–d) Comparison of recorded data from 30-μm diameter Ti, Pt, and PEDOT:PSS electrode contacts, sampled at 20 kHz. The three contacts are placed spatially adjacent to each other, as shown in the inset of panel (c). (a) Baseline noise recorded from Ti, Pt, and PEDOT:PSS electrode contacts in vivo, high-pass filtered at 300 Hz, and the corresponding root mean square (RMS) noise recorded on each material for a 20-ms recording. (b) Filtered high-gamma activity (70–190 Hz) recorded in vivo on the barrel cortex of a rat in response to an air puff stimulation applied on the whisker. The data show the response plotted for multiple trials, with the average trial-averaged waveform plotted in bold for each material. (c) The corresponding signal-to-noise ratio (SNR) of the three materials extracted from the results plotted in (a) and (b). (d) The trial-averaged response measured on each material, comparing the relative variation in the maximum amplitude measured on each material, aligned to the onset of air stimulation. The inset shows the average delay in the positive peak of the response on the Pt (89.5 μs) and Ti (268 μs) contacts, with respect to the PEDOT contact. (e) Optical and electron microscope images of nanowire, microwire, and millimeter wire electrodes used to study the variation of the (f) 1 kHz impedance, in benchtop measurements, for different electrode contact materials. (g) The variation of the coupling coefficient as a function of the equivalent R and C at the electrochemical interface.

Figure 3.

Recorded activity using a high-density, high channel count PtNRGrid placed on the surface of a human brain during a craniotomy. (a) Overlay plot of high-gamma activity sensory responses superimposed on top of a photo of the surface of a patient’s brain, in response to vibrotactile stimulation of individual fingers of the patient. Functional boundary (FB). (b–d) The patient is asked to perform a grasping task using the hand, and the measured propagating beta waves and waveforms are plotted across the central sulcus (CS) in the (b) planning stage of the motion, (c) during the motion, and (d) after the completion of motion of a patient’s hand. The red and blue streamlines originate from the sensory (S) and motor (M) cortices, respectively. The background color represents the amplitude of the beta wave potential, and the arrowheads indicate the propagating direction of the beta waves. Bottom plots are raw waveforms around the time stamps of (b) to (d), with the arrowheads indicating the propagation direction of the beta waves.

Figure 4.

Direct electrical stimulation via intracranial electrodes to drive neural activity and the effects of electrode material and size on the measured impedance spectra. (a) The equivalent circuit model for current injection in vivo, showing the individual elements of the electrode–tissue interface that participate in the charge-injection process individually delineated. (b) Material- and (c) diameter-dependent electrical impedance spectra (EIS) in benchtop measurements. The diameter dependence of the (d) double layer impedance, (e) charge-transfer resistance, and (f) faradic impedance. (g) The bias-dependent variation of the electrochemical interface elements in benchtop.

Figure 5.

Dependence of maximum injectable current on stimulation design parameters. Maximum injectable current as a function of (a) pulse width for 200-μm PtNR, PEDOT:PSS, and planar Pt contacts, (b) diameter for a 200-μs pulse for in vivo (rat) and benchtop placement, (c) pulse width for a 200-μm PtNR contact for different inter-contact separations.

Figure 6.

(a) The agreement between the modeled and simulated data for the cathodal excitation as a function of the pulse width and current. (b) The agreement between the modeled and simulated data for the cathodal excitation as a function of the current and impedance. (c) The safety limits predicted by the predictive equation compared to the limits proposed by Shannon’s Equation. (d–f) The fitting results for the cathodal excitation measured on the pig’s cortex using clinical electrodes, plotted as a function of the input current and pulse width, for (d) depth, (e) stereo-encephalography (sEEG), and (f) surface strip electrodes.

Figure 7.

Connectorization to clinical and high channel count electrodes. (a) Example of commonly used connectors for current clinical electrodes. Each channel on these electrodes is individually routed to leads, which makes this approach unscalable for high channel counts. (b) Currently clinically adopted connector technologies developed by Blackrock Neurotech, for human implants using the Utah Microelectrode Arrays (Cereplex I-128 Implantable Electrode and Amplifier—© 2023 Blackrock Neurotech, LLC). (c) University of California, San Diego’s (UCSD) PtNRGrids use conventional land grid array (LGA)-type connectors and custom-built high-density Ironwood connectors to bond the flexible electrode to an extender printed circuit board to allow high-density connections to acquisition electronics. (d) IMEC Neuropixels provides 960 channels for recording and uses an integrated connector with a complementary metal oxide semiconductor (CMOS) digital neural probe integrated to the on-chip circuitry. (e) Paradromics Argo’s integrated connector vertically bonds interconnects on rigid substrates, allowing ultrahigh-density multi-thousand channel devices. (f) Flip-chip bonding technique developed at the University of Michigan–Ann Arbor using anisotropic conductive film (ACF) bonding technique to bond flexible substrates to CMOS chips, yielding a connection density of 167 channels/mm². LNA, low-noise amplifier.

Figure 8.

Schematic of a human-grade high-density, high channel count electrode array with a wireless acquisition system with simultaneous recording and stimulation capabilities for chronic implants, currently under development by the present authors and collaborators.