Recent work has demonstrated the feasibility of minimally-invasive implantation of electrodes into a cortical blood vessel. However, the effect of the dura and blood vessel on recording signal quality is not understood and may be a critical factor impacting implementation of a closed-loop endovascular neuromodulation system. The present work compares the performance and recording signal quality of a minimally-invasive endovascular neural interface with conventional subdural and epidural interfaces. We compared bandwidth, signal-to-noise ratio, and spatial resolution of recorded cortical signals using subdural, epidural and endovascular arrays four weeks after implantation in sheep. We show that the quality of the signals (bandwidth and signal-to-noise ratio) of the endovascular neural interface is not significantly different from conventional neural sensors. However, the spatial resolution depends on the array location and the frequency of recording. We also show that there is a direct correlation between the signal-noise-ratio and classification accuracy, and that decoding accuracy is comparable between electrode arrays. These results support the consideration for use of an endovascular neural interface in a clinical trial of a novel closed-loop neuromodulation technology.

Objective: Subdural electrocorticography (ECoG) can provide a robust control signal for a brain-computer interface (BCI). However, the long-term recording properties of ECoG are poorly understood as most ECoG studies in the BCI field have only used signals recorded for less than 28 days. We assessed human ECoG recordings over durations of many months to investigate changes to recording quality that occur with long-term implantation. Methods: We examined changes in signal measurements properties over time from 15 ambulatory humans who had continuous subdural ECoG monitoring for 184 to 766 days. Results: Individual electrodes demonstrated varying changes in frequency power characteristics over time within individual patients and between patients. Group level analyses demonstrated that there were only small changes in effective signal bandwidth and spectral band power across months. High-gamma signals could be recorded throughout the study, though there was a decline in signal power for some electrodes. Conclusion: ECoG-based BCI systems can robustly record high-frequency activity over multiple years, albeit with marked inter-subject variability. Significance: Group level results demonstrated that ECoG is a promising modality for long-term BCI and neural prosthesis applications.

Objective. Recently, we reported a minimally invasive stent-electrode array capable of recording neural signals from within a blood vessel. We now investigate the use of electrochemical impedance spectroscopy (EIS) measurements to infer changes occurring to the electrode–tissue interface from devices implanted in a cohort of sheep for up to 190 days. Approach. In a cohort of 15 sheep, endovascular stent-electrode arrays were implanted in the superior sagittal sinus overlying the motor cortex for up to 190 days. EIS was performed routinely to quantify viable electrodes for up to 91 days. An equivalent circuit model (ECM) was developed from the in vivo measurements to characterize the electrode–tissue interface changes occurring to the electrodes chronically implanted within a blood vessel. Post-mortem histological assessment of stent and electrode incorporation into the wall of the cortical vessels was compared to the electrical impedance measurements. Main results. EIS could be used to infer electrode viability and was consistent with x-ray analysis performed in vivo , and post-mortem evaluation. Viable electrodes exhibited consistent 1 kHz impedances across the 91 day measurement period, with the peak resistance frequency for the acquired data also stable over time. There was a significant change in 100 Hz phase angles, increasing from −67.8° ± 8.8° at day 0 to −43.8° ± 0.8° at day 91, which was observed to stabilize after eight days. ECM’s modeled to the data suggested this change was due to an increase in the capacitance of the electrode–tissue interface. This was supported by histological assessment with >85% of the implanted stent struts covered with neointima and incorporated into the blood vessel within two weeks. Conclusion. This work demonstrated that EIS could be used to determine the viability of electrode implanted chronically within a blood vessel. Impedance measurements alone were not observed to be a useful predictor of alterations occurring at the electrode tissue interface. However, measurement of 100 Hz phase angles was in good agreement with the capacitive changes predicted by the ECM and consistent with suggestions that this represents protein absorption on the electrode surface. 100 Hz phase angles stabilized after 8 days, consistent with histologically assessed samples. Significance. These findings demonstrate the potential application of this technology for use as a chronic neural recording system and indicate the importance of conducting EIS as a measure to identify viable electrodes and changes occurring at the electrode–tissue interface.

Electrodes are used in vivo for chemical sensing, electrophysiological recording, and stimulation of tissue. The electrode configuration used in vivo is often optimised for a specific anatomy and biological or clinical outcomes, not electrochemical performance. Electrode materials and geometries are constrained by biostability and biocompatibility issues and may be required to function clinically for decades. We performed benchtop electrochemistry, with changes in reference electrode, smaller counter-electrode sizes, and three- or two-electrode configurations. We detail the effects different electrode configurations have on typical electroanalytical techniques used on implanted electrodes. Changes in reference electrode required correction by application of an offset potential. In a two-electrode configuration with similar working and reference/counter-electrode sizes, the electrochemical response was dictated by the rate-limiting charge transfer step at either electrode. This could invalidate calibration curves, standard analytical methods, and equations, and prevent use of commercial simulation software. We provide methods for determining if an electrode configuration is affecting the in vivo electrochemical response. We recommend sufficient details be provided in experimental sections on electronics, electrode configuration, and their calibration to justify results and discussion. In conclusion, the experimental limitations of performing in vivo electrochemistry may dictate what types of measurements and analyses are possible, such as obtaining relative rather than absolute measurements.