Development of a novel, concentric micro-ECoG array enabling simultaneous detection of a single location by multiple electrode sizes

Objective. Detection of the epileptogenic zone is critical, especially for patients with drug-resistant epilepsy. Accurately mapping cortical regions exhibiting high activity during spontaneous seizure events while detecting neural activity up to 500 Hz can assist clinicians’ surgical decisions and improve patient outcomes. Approach. We designed, fabricated, and tested a novel hybrid, multi-scale micro-electrocorticography (micro-ECoG) array with a unique embedded configuration. This array was compared to a commercially available microelectrode array (Neuronexus) for recording neural activity in rodent sensory cortex elicited by somatosensory evoked potentials and pilocarpine-induced seizures. Main results. Evoked potentials and spatial maps recorded by the multi-scale array (‘micros’, ‘mesos’, and ‘macros’ refering to the relative electrode sizes, 40 micron, 1 mm, and 4 mm respectively) were comparable to the Neuronexus array. The SSEPs recorded with the micros had higher peak amplitudes and greater signal power than those recorded by the larger mesos and macro. Seizure onset events and high-frequency oscillations (∼450 Hz) were detected on the multi-scale, similar to the commercially available array. The micros had greater SNR than the mesos and macro over the 5-1000 Hz frequency range during seizure monitoring. During cortical stimulation experimentation, the mesos successfully elicited motor effects. Significance. Previous studies have compared macro- and microelectrodes for localizing seizure activity in adjacent regions. The multi-scale design validated here is the first to simultaneously measure macro- and microelectrode signals from the same overlapping cortical area. This enables direct comparison of microelectrode recordings to the macroelectrode recordings used in standard neurosurgical practice. Previous studies have also shown that cortical regions generating high-frequency oscillations are at an increased risk for becoming epileptogenic zones. More accurate mapping of these micro seizures may improve surgical outcomes for epilepsy patients.


Introduction
The development and evolution of brain-machine interfaces (BMIs) capable of providing high-fidelity information on neurological signals has significantly improved the diagnosis and treatment of neurological disorders.Amongst the sea of neurological diseases yet to be fully understood, epilepsy has been at the forefront of research attempting to find ways of detecting seizure events and, hopefully, one day, eliminate epilepsy (Tiwari et al 2019).Patients with epilepsy make up nearly 1% of the worlds population, and of that, roughly 30% fall under the drug-resistant category despite the advances in anti-seizure medication (Englot & Chang 2014, Chen et al 2018).Surgical resection has been viewed as one of the more effective options for patients with drug-resistant, focal epilepsy even though it is typically effective in providing patients with long-term seizure freedom 50% of the time (Malmgren & Edelvik 2017, Landazuri et al 2020).Inherent risks of surgical intervention include complications during surgery, the unnecessary removal of neural tissue, and, of course, the risk of recurrence.Neuromodulation therapies that have been approved as alternatives to medication and antiseizure medication, such as vagus nerve stimulation (VNS), deep brain stimulation (DBS), and closed-loop responsive neurostimulation (RNS), have provided temporary relief but seldom long-lasting freedom (Ryvlin et al 2021).
For surgical resection or laser ablation to be effective, proper determination and targeting of the epileptogenic zone (EZ), or brain region responsible for generating seizures, is necessary (Jehi 2018).Diagnostic tools used to identify the EZ include scalp electroencephalography (EEG), magnetic resonance imaging (MRI), and intracranial EEG (iEEG) techniques: stereo encephalography (sEEG), and electrocorticography (ECoG) (Hasegawa 2016).Implantable probes such as thin-film polymer-based microelectrode arrays (MEAs) are specific iEEG BMIs that play a significant role in our understanding of what is happening intracranially during seizures.Because implantation of these MEAs can cause severe lesions of the target tissue, advances in the design and development of minimally invasive micro-ECoG arrays have received increased attention (Li et al 2021, Vomero et al 2020, Shokoueinejad et al 2019, Lee et al 2023).Biocompatibility, flexibility, conductivity, and elasticity are the primary design considerations of thin-film MEAs because of the environment that is the surface of the brain.A material mismatch between the surrounding neural tissue and the implantable device is one of the leading causes of device failure and poor recording efficacy.Commonly used polymers in the design of thin-film MEAs are polyimide (PI), parylene-C, and polydimethylsiloxane (PDMS) because of their reduced Youngs modulus.
Traditional monitoring of epileptic activity was typically focused over the Berger bands (1-25 Hz) using widely spaced clinical macro-sized electrodes.However, studies exploring the power spectral density (PSD) of neural activity in the higher frequency ranges of gamma (25-80 Hz), ripple (80-200 Hz), and fast ripple (>250 Hz) have shown these frequencies also show changes in the epileptic brain (Worrell et al 2008, 2012, Park & Hong 2019, Schönberger et al 2020).Recordings from micro-ECoG arrays on epileptic patients have shown that spatially localized epileptiform activity at higher frequencies provide helpful information that macro electrodes are incapable of recording.The inclusion of macro electrodes and microelectrodes on a single array, called hybrid electrode arrays, demonstrated this advantage ( Van Gompel et al 2008, Stead et al 2010).Macro and microelectrodes co-existed on the same arrays with the microelectrodes incorporated into a standard macro electrode array and showed that the detection of micro seizures on the microelectrodes were potential interictal biomarkers of epileptic tissue.Because of the overall design of those arrays, it remains to be seen if the localization of focal seizures is possible through the combined abilities of macro and micro-sized electrodes within the same recording space.
In this work, we leveraged the properties of micro-ECoG electrode arrays and the hybrid array concepts described previously to develop a novel 32-channel, multi-scale micro-ECoG electrode array.The 32channel multi-scale concept incorporated three electrode sizes (40 μm, 1 mm, and 4 mm) in a novel concentric design where we carried out in vitro tests and in vivo recordings of evoked potentials over the somatosensory cortex while also detecting clinical and micro seizure activity.Our objectives were to assess if this novel design was capable of spatially mapping somatosensory evoked potentials (SSEPs) comparably to a commercially available micro-ECoG electrode array (Neuronexus Technologies), record pilocarpineinduced seizure events, and analyze which of the three electrode sizes within the hybrid array design provides the greatest signal-to-noise ratio (SNR).

Design of multi-scale array
Electrode geometries for the concentric multi-scale micro-ECoG array (figure 1(A)) were designed to include electrode sizes similar to existing clinical standard electrodes in order to directly compare with conventional devices.The 4 mm macro and 40 μm micros correlate well to the typical diameters of subdural grid electrodes and strips as well as platinum microwires.The 1 mm mesos were intended to mimic commonly found penetrating electrodes that have 1 mm diameters.The inclusion of the macro and meso electrode sizes were also influenced by the potential to stimulate and invoke seizure activity locally.Based on a suggestion from Worrell et al (2012), center-tocenter electrode spacing of 350 μm for the micros and 1.5 mm for mesos were implemented for complete coverage without overlap.The concentric configuration was designed to increase the electrode density while reducing the number of connections enabling enhanced spatial detail where desired and regional summation throughout.In total, the multi-scale array is composed of 1 macroelectrode, 4 mesoelectrodes, and 27 microelectrodes (figure 1(C)).

Fabrication of multi-scale array
A polyimide substrate was chosen to create a thin, highly flexible micro-ECoG array with a simple manufacturing process.Polyimide is a widely used material for creating flexible printed circuit boards (PCBs) (Lenihan et al 1996, Engel et al 2003, Dobrzynska & Gijs 2012).A spin-coated polyimide variant, PI-2611 (HD Microsystems, Parlin, NJ, USA), was used to create a 10μm substrate layer and 10 μm insulation layer for a total thickness of 20 μm which optimizes the array's flexibility.Electrodes were created by depositing 5 nm of titanium (Ti) and 100 nm of platinum (Pt), which enabled a stable charge density and ensured biocompatibility for neural recording and stimulation.Figure 1(B) highlights the fabrication process step-by-step.Metal adhesion was confirmed with an ASTM D3359 tape test.Contacts were exposed by dry etching PI-2611 with oxygen plasma to prevent biotoxicity.Pieces of 127 μm thickness polyether ether ketone (PEEK) were applied as backing to insert the array into a 39-pin zero insertion force (ZIF) connector.A custom electrode interface board (EIB) design enabled connection from the ZIF connector to an Intan RHD electrophysiology evaluation system (Intan Technologies, Los Angeles, California, USA) through Omnetics connectors (Omnetics Corporation, Minneapolis, MN, USA).Additional EIB designs were created to interface the multi-scale array and benchtop characterization equipment.All connections were confirmed with multimeter probes before use.

Recording of somatosensory evoked potentials
Three (two male, one female) Sprague-Dawley rats (250-350g) were used in gathering SSEP data.To prepare for measurements, rats were placed in a stereotactic apparatus and anesthetized using an initial dose of 5% isoflurane and then maintained under anesthesia with a mix of 2% isoflurane and continuous oxygen flowing at 2 liters per minute (lpm) delivered through a nose cone.Vitals were monitored with a MouseOx Plus pulse oximeter (Starr Life Sciences Corporation, Oakmont, PA, USA), and internal temperature was maintained with a heating pad and towels and monitored by a rectal probe.A single 5 mm x 10 mm window craniotomy was performed over the left hemisphere (coordinates with respect to bregma; 7 mm posterior, 3 mm anterior, 5.5 mm lateral, 0.5 mm medial).Before stimulation, sites were shaved and cleaned with disinfectant.A similar approach to previous studies was taken in the gathering and processing of SSEP data (Hosp et al 2008, Moncion et al 2022).Constant current pulses (1Hz, 0.5 mA, 2 ms duration) were delivered via bipolar needle electrodes to the median and tibialis nerves (A-M Systems, Sequim, WA, USA).Epidural potentials were amplified through an RHD 2132 32-channel headstage (Intan, Los Angeles, CA, USA), obtained with a 5 kHz sampling rate and filtered (0.72-7000 Hz bandpass).
An illustrated schematic showing an overview of the recording setup is shown in figure 3(A) 2.6.Pilocarpine induced epileptic rat model Four (two male, two female) Sprague-Dawley rats (250-350g) were used as seizure models for this study.
Rats were anesthetized using an initial dose of 5% isoflurane and then maintained under anesthesia with a mix of 2% isoflurane and continuous oxygen flowing at 2 liters per minute (lpm) through a nose cone.A double window craniotomy was performed, with each window being 5 mm x 7 mm (figure 4).Baseline neural signals were recorded for 5 minutes to determine regular activity.Following baseline recording, 300 mg/kg of pilocarpine hydrochloride (Sigma Aldrich, St. Louis, MO, USA) was injected intraperitoneally following similar procedures to previous studies (Curia et al 2008, Turski et al 1983).Neural signals were recorded with a 5 kHz sampling frequency using two RHD 2132 32-channel headstages.Intracranial recording commenced, and we maintained visual observation of potential seizure symptoms.

Cortical stimulation evaluation
Two (one male, one female) Sprague-Dawley rats (250-350g) were used.Instead of isoflurane, a ketamine-xylazine anesthetic was chosen to preserve motor nerve function.Before the craniotomy, anesthesia was induced using the ketamine / xylazine (KX) cocktail with a dose of (75/10 mg/kg) through intraperitoneal (IP) injection.Anesthesia was maintained using a ketamine dose of 25 mg/kg administered every 45 to 60 minutes and a xylazine dose of 5 mg/kg every 2 hours.Anesthesia was verified by squeezing the rat's foot pad every 15 minutes to confirm the loss of withdrawal response.A window craniotomy was performed 4 mm posterior to lambda and 4 mm anterior to bregma, and 4 mm lateral from midline for an entire 4 mm × 12 mm window.A 200 μm stainless steel wire was placed under the scalp as a reference wire.Cortical stimulation was performed across two meso electrode configurations (between electrodes m2 and m16 and between electrodes m16 and m17, as shown by the red and green double arrows) shown in figure 5. Placement 1 and placement 2 depict the two locations on the cortical surface that were stimulated using an isolated pulse stimulator (Model 2100, A-M Systems, Sequim, WA,   activity.This method of SNR quantification is referred to as the voltage SNR, or vSNR and is broadly used when there is not a need for the full information provided by spectral analysis (Blaschke et al 2017).However, because the signals of interest have frequency dependent information encoded, we computed the spectral SNR.Spectral SNR was computed as the ratio of the power spectral density (PSD) during the active state with respect to the power during the resting state (Suarez-Perez et al 2018).In this study, we investigated the spectral SNR of the micro, meso, and macro electrodes on the multi-scale array to determine the recording efficacy of the varying electrode sizes at different relevant frequencies.

Electrochemical impedance spectroscopy
Complex impedances were measured from 54 micros, 8 mesos, and 2 macro electrodes on the multi-scale array (2 arrays) and all 32 electrodes from the NNx array.The measurements were consistent throughout the recordings.Impedances shown in figure 1(D) were as expected for electrode geometries on the multi-scale array; lower impedances for larger electrodes and higher impedances for smaller electrodes.The impedance magnitudes at 1 kHz were: 3.2 ± 0.2 kΩ for the macros, 4.4 ± 0.4 kΩ for the mesos, 212.2 ± 20.4 kΩ for the micros, and 105.5 ± 12.3 kΩ for the NNx electrodes.

Somatosensory evoked activity recording
Somatosensory evoked potentials were chosen as the first in vivo experiments because they are widely accepted in the research community with commonly known characteristics and allowed for direct comparison of spatial resolution between our multi-scale array and the NNx array.The SSEP's distinct timelocked waveform can be expected and predicted from a site-specific stimulus.The recorded evoked potentials were averaged over multiple stimuli events to create an average waveform representative of the elicited activity.
In this study, we monitored evoked responses from hindlimb (HL) and forelimb (FL) stimulation.figure 6(A) shows the response measured from the HL stimulation, and figure 6(B) shows the response measured from the FL stimulation for the NNx and multiscale arrays.Above each SSEP trace is a heat map indicating the spatial distribution of the evoked activity at varying time points along the SSEP trace to determine localization.The HL heat map shows more pronounced cortical localization of the evoked response than the FL heat map, which we attribute to centering the arrays closer to the HL region than the FL region of the somatosensory cortex.Although the electrode sizes from both arrays allow for adequate spatial coverage, the overall array size does not allow for full coverage of the somatosensory region of the cortex.Figures 6(A) and 6(B) also highlight the differences between the recorded signal from the NNx array and our multi-scale array for HL stimulation and FL stimulation.Most traces on the NNx array appear to have greater peak amplitudes than the peaks of the multi-scale array, which we ascribe to the larger electrode size of the NNx array (100 μm) compared to the micros on the multi-scale array (40 μm).Average latency and FWHM were plotted in figures 6(C) and 6(D) for both arrays.The NNx SSEP HL stimulation had an average peak latency of 18.40 ± 1.56 ms and FWHM of 17.89 ± 1.98 ms.The corresponding values for the multi-scale array were 21.11 ± 2.11 ms and 15.66 ± 2.72 ms (figure 6(C)).Results obtained from both arrays for HL SSEP were highly consistent.The FL SSEP recordings for the NNx array resulted in an average peak latency and FWHM of 12.82 ± 1.44 ms and 15.43 ± 2.31 ms, respectively.The corresponding values for the multi-scale array were 13.63 ± 1.95 ms and 14.65 ± 2.65 ms (figure 6(D)).Analysis using the non-parametric Mann-Whitney test (p < 0.05) revealed that none of the values between the multiscale array and NNx array were statistically significant.
To examine the recording capability of the different electrode sizes, a deeper analysis of the recorded traces from the micro, meso, and macro electrodes were investigated from the HL stimulation.Figure 7(A) indicates which set of electrodes are analyzed in 7(B).The average of the micro electrodes in the quadrant boxed in red are compared to the meso electrode from that same quadrant.The average of the micro traces had a peak amplitude of 29.18 ± 3.24 μV and a peak latency of 20.8 ± 1.2 ms.The meso electrode from that same quadrant had a peak amplitude of 28.33 ± 1.26 μV and peak latency of 21.4 ±0.76 ms.The bottom plot in figure 7(B) highlights the average traces of the meso electrodes compared to the activity trace of the macro electrode from the same HL stimulation.The averaged mesos had a peak amplitude of 18.36 ± 4.34 μV and peak latency of 21.2 ± 0.95 ms, while the macro had a peak amplitude of 11.65 ±1.87 μV and peak latency of 21.4 ± 0.34 ms.The differences in peak latencies and FWHM were all statistically insignificant.Figure 7(C) illustrates the difference in spatial mapping when only the macro electrode is used (left), only four meso electrodes are used (middle) and only the 27 micro electrodes are used (right).The heat map from the micro only distribution clearly showed an improved spatial resolution of the high activity localization when compared to the meso only heat map.
We have demonstrated that the hybrid, multiscale array can record neuronal modulation and performs comparably to a commercially available micro-ECoG array.The performance of the micro, meso, and   macro electrodes were compared to one another during the SSEP validation and results clearly were indicative that the micros had improved signal power and localization of the evoked activity on the cortex.

Seizure model
Epileptiform discharges and seizure propagation were simultaneously monitored on both hemispheres by completing a double craniotomy and placing the multi-scale micro-ECoG array on one hemisphere and the NNx on the other.The arrays are centered over the same locations, covering multiple cortical regions on each hemisphere.Figure 8 shows selected examples of seizure events recorded from both arrays.Figure 8(A) depicts the transition from a normal, non-epileptic state to a seizure state from the multi-scale array.Cortical electrophysiological activity intensifies significantly during the seizure onset.The peak-to-peak potentials during the baseline state were 99.23 ± 1.09 μV and increased to 492.17 ± 7.01 μV during the seizure state.The power spectral density (PSD) for all 32 electrodes on the multi-scale array was calculated during baseline and seizure status (figure 8(B)).The PSD curves exhibited the greatest concentration of power in the 1-30 Hz region, with the curves during the seizure status showing higher power than during the baseline phase.Significant peaks occurred in the 5-7 Hz theta frequency band, a hallmark of seizure activity.Studies show that change in theta band power may be a biomarker for potential epileptogenesis.As shown in the HL SSEP analysis, the micros clearly showed improved signal power over a large frequency range (5-1000 Hz) when compared tp the meso and macro electrodes during the seizure state (figure 8(C)).The quadrant of electrodes discussed in figure 7(A) are illustrated again in figure 8(D) and show a recorded interictal spike (red rectangle).The example of this interictal spike shows the difference in amplitude between similarly located micros and their respective meso electrode.
The microelectrodes on the multi-scale array were also capable of recording high-frequency oscillations (HFOs), as expected due to the appropriately small size of the recording electrodes (<100 μm).segment of the recorded HFO from a micro (top), meso (middle), and macro (bottom) electrode and clearly highlights the diminishing strength of the HFO as the size of the recording electrode increased.

Spectral SNR
Electrode SNR is dependent on electrode impedance and thus makes it frequency-dependent.Because of this, quantitatively evaluating SNR at different biological frequencies is essential in characterizing electrode performance.Spectral SNR was evaluated for each electrode size (micro, meso, and macro) in order to compare them over the biologically relevant frequency ranges by calculating the PSD of the Up states (signal) and the Down states (baseline) (figure 10(A)).Overall, from the spectral SNR analysis we found that all three electrode sizes had relatively consistent SNR in the low frequency range up to 30 Hz (figure 10(B).At this frequency range, the micro electrodes had an SNR roughly about 17 dB and the meso electrode around 16 dB.The SNR of the macro in this frequency range was significantly lower at about 9 dB.For frequencies greater than 30 Hz, the SNR decayed almost linearly following the typical 1/f decay until each electrode hit a plateau at varying frequencies.Throughout the entire frequency spectrum analyzed the electrode SNR was inversely related to the electrode size, meaning the micros had the greatest SNR and the macros had the smallest.These results corroborated what was seen in the SSEP analysis (figure 8(D)).The micro electrodes were the only electrode size to have positive SNR throughout the entire low and medium frequency ranges and eventually plateaued at 0 dB past 700 Hz while the meso and macro electrodes eventually reached negative SNR past 100 Hz and 85 Hz respectively.Negative SNR at the higher frequencies were caused because the Down PSD states exceeded the Up PSD states.This effect may be caused by some noise artifact.As expected, the recording performance of the electrodes diminished at higher frequencies, with the micros performing best.

Cortical stimulation
Stimulus pulse widths of 100 μs were tested first but did not produce successful motor effects given the current limit, so 200 μs pulses were used instead.Table 1 reports the minimal cortical threshold current (in mA) required to elicit an observable movement from the two separate array placements on the brain.Forelimb activation occurred at a minimum cortical simulation of 2.6 mA between the m16 and m17 meso electrodes.The stimulation between the m2 and m16 meso electrodes, which are laterally aligned, evoked whisker movement and wrist pronation at a stimulus of 3.2 mA.In an attempt to seek additional motor mapping, the multi-scale array was moved 2 mm caudally, and the stimulation patterns were repeated.Hindlimb activation was observed from a 2.4 mA stimulation between the m16 and m17 electrodes.The stimulation between the m2 and m16 electrodes elicited forelimb and jaw movement at a stimulus of 3.6 mA.

Discussion
In this study, we successfully designed, fabricated, and characterized a flexible, multi-scale, hybrid micro-ECoG array for recording SSEP neural activity and pilocarpine-induced seizure activity.As expected, the electrochemical characterization showed that the impedance magnitudes of the multi-scale array decreased with increasing electrode size.Successful recording and spatial mapping of the hindlimb and forelimb SSEPs highlighted the similarity in spatial resolution between the 'micros' on the multi-scale array and the electrodes on the commercially available NNx array.Although the recorded evoked potentials from the micros on the multi-scale were slightly diminished compared to those recorded using the electrodes on the NNx array, we attribute this to the difference in their electrode size.Simultaneous monitoring of the SSEPs from the multiple electrode sizes on the multi-scale array showed that the micros had increased signal power and improved spatial localization of elicited activity over the mesos and macros.
Our results also verified what others had hypothesized: SSEP amplitudes were greater when recorded from smaller electrodes (Castagnola et al 2015).
The model of epilepsy induced by pilocarpine effectively produced electrographic seizures and HFOs that are typical of epileptogenic tissue.These HFOs have been linked to the regions that generate seizures and the development of chronic epileptogenesis, as demonstrated by previous studies (Bragin et al 2000, Schönberger et al 2020).Our study aimed to address the hypothesis put forward by Sindhu et al (2023), who altered the physical area of an electrode array by shorting neighboring electrodes.We monitored activity simultaneously using multiple electrode sizes in the same location, unlike previous studies that could only sequentially monitor the same area with electrodes of various sizes.Our results showed that microelectrodes were better at recording activity with higher signal power and improved SNR, as measured using spectral SNR analysis.This analysis provided frequencydependent insights relevant to the neurologically significant frequency ranges.Interestingly, we found that the smallest electrodes (micros) had the highest SNR, while the largest electrode (macro) had the lowest SNR from 5-1000Hz (figure 10).Furthermore, compared to mesos and macro electrodes, micros were significantly more effective at detecting HFOs (figure 9).
However, our study had a few limitations.Although the novel concentric design allowed us to monitor the same location with multiple electrode sizes simultaneously, it limited the ability to produce traditional heat maps and compare the spatial resolution against a standard grid array.Additionally, clinical translation of the multi-scale array would require extensive animal testing, long-term studies, and scaled manufacturing capabilities.Multi-electrode arrays with polyimide substrates have been proven successful in long-term animal use when implanted for over six months, with the only issue being dura mater growth encompassing the array (Romanelli et al 2019).In parallel with developing new bio-materials to prevent biofouling of the implanted array, a need for improved in vitro and ex vivo models before trials move to in vivo models would have the potential to reduce the total number of animal models significantly used (Gulino et al 2019).
This innovative micro-ECoG array is the first of its kind to enable the simultaneous monitoring of multiple electrode sizes from a single location in the brain.By examining neurological phenomena across various electrode sizes, we can gain a better understanding of how electrode size impacts the quality and types of neurological events recorded.For years, clinicians and researchers have been searching for effective ways to localize the epileptogenic region in patients with drugresistant epilepsy, and the detection of HFOs has emerged as a crucial indicator (Jacobs & Zijlmans 2020).The groundbreaking research of Morrell (2011) has shown that seizures can be significantly reduced through cortical stimulation of predetermined seizure foci.Our multi-scale array design not only allows us to capture brain activity, but also to stimulate it, providing clinicians with multiple treatment options for patients with drug-resistant epilepsy.By leveraging the multi-scale array to detect HFOs with the micros and then employing the mesos for stimulation, our work paves the way for significant progress in the treatment of drug-resistant epilepsy.

Conclusion
Our research has successfully analyzed the trade-offs between the multiple electrode sizes of the concentric multi-scale array by monitoring a range of neurological signals elicited from SSEPs and pilocarpineinduced seizures.We have thoroughly investigated the impact of electrode size over equivalent recording areas and found that using a clinically-sized macro electrode does not increase the SNR at lower frequencies as hypothesized.Moreover, our analysis has revealed that using larger electrodes negatively impacts spatial resolution in areas of interest.Our results highlight the unique benefits of high-density microelectrode arrays, which offer exceptional spatial resolution and superior SNR of neurological signals across clinically relevant frequency ranges.Although minute changes in overall electrode size may have frequencydependent effects, our data unequivocally show that using uniformly sized micro-electrode arrays is more than sufficient for monitoring neural signals.The experimental analysis performed with the multi-scale array clearly demonstrated that the micro electrodes out performed the larger meso and macro electrodes in SNR and spatial resolution, thus paving the way for future work to focus on using high-density microelectrode interfaces.

Figure 1 .
Figure 1.32-channel multi-scale array: A. Picture of array after release from wafer.B. Protocol used to fabricate multi-scale array: (a) silicon wafer; (b) spin polyimide; (c) metal layer deposition; (d) patterning; (e) insulation layer deposition and spin photoresist; (f) pattern, etch and removal of photoresist and release from silicon wafer.C. Microscope image of 32-channel multi-scale array with electrodes highlighted (micros: red, mesos: blue, and macro: green).D. Electrochemical impedance spectroscopy (EIS) magnitude plot of micros, mesos, and macro.

Figure 2 .
Figure 2. Multi-scale and Neuronexus arrays.A. Magnified image of the multi-scale and Neuronexus arrays detailing the electrodes.B. Placement location of the arrays over the somatosensory cortex; S1HL: hindlimb, S1FL: forelimb.

Figure 3 .
Figure 3. SSEP experimental setup.A. Overview of the experimental SSEP setup indicating the stimulation, signal pathway, and expected recorded signal.B. SSEP recording overview highlighting the transition from the recording of the raw signal, followed by transposing the raw signal into 1 second windows, and lastly averaging those windows to create the final SSEP trace.

Figure 4 .
Figure 4. Diagram indicating the double craniotomy and placement of the multi-scale and NNx arrays (left) with actual image from one surgery (right).

Figure 5 .
Figure 5. Schematic of multi-scale array placements for cortical stimulation.Placements 1 and 2 depict the two array placements during the cortical stimulation and are located approximately 2 mm apart.The m2/m16 and m16/m17 arrows highlight the stimulation direction for each stimulation pair.

Figure 6 .
Figure 6.Somatosensory evoked activity from hindlimb (HL) and forelimb (FL) stimulation.A. SSEP HL activity traces from all 32 electrodes on the NNx (blue) and multi-scale (black) array with the red vertical line indicating the stimulation point.Heat maps show the spatial distribution of the SSEP activity corresponding to the array at each time point denoted by the dashed gray lines.(B.) SSEP FL activity traces from all 32 electrodes on the NNx (blue) and multi-scale (black) array with the red vertical line indicating the stimulation point.Heat maps show the spatial distribution of SSEP activity corresponding to the array at each time point denoted by the dashed gray lines.(C.) Bar chart comparing SSEP HL latency and FWHM from the average of the traces from each array.(D.) Bar chart comparing SSEP FL latency and FWHM from the average of the traces from each array.

Figure 7 .
Figure 7. Somatosensory evoked activity from hindlimb (HL) stimulation.(A.) Magnified image of multi-scale array highlighting the specific electrodes that are analyzed in B. (B.) Top: Data from a quadrant of electrodes (red box) that compares the average recorded activity of the micros (red) in that quadrant to the recorded activity of the meso (blue) from 0.5 mA HL stimulation.Bottom: Data from 0.5 mA HL stimulation comparing the average recorded activity of the four mesos (blue) to the recorded activity of the macro (green).(C.)Heat map indicating the spatial resolution from HL stimulation from left: the macro electrode only, middle: the meso electrodes only, and right: the micro electrodes only.

Figure 8 .
Figure 8. Epileptic events recorded from multi-scale array.(A.) ECoG signal change recorded on an example subset of electrodes on the multi-scale array during a transition from baseline to seizure status.(B.)Power spectral density comparing the baseline and seizure status.(C.)Average power spectra of the micros, mesos, and macro from 5-1000 Hz during the seizure state.(D.) Example of an interictal seizure spike (red rectangle) from the micros and meso from the same quadrant on the multi-scale array.
Figure 9(A) and (B) show an example of a 450 Hz signal recorded from both arrays depicting the asynchronous and spatially diverse nature of the recorded HFO across nonadjacent microelectrodes.The shaded lines in figure 9(C) depict the recorded signal from each individual electrode on the multi-scale array, while the bolded line depicts the average for each respective electrode geometrical size (micro, meso, macro) for the recorded HFO segment highlighted in figure 9(B).The spectrogram in figure 9(D) shows the time

Figure 9 .
Figure 9. HFO recordings (A.) HFO recorded from a subset of electrodes on the NNx array.(B.) Same HFO recorded from a subset of electrodes on the novel multi-scale array.(C.)Same HFO recording from the plots above; bold traces highlight the average for the micros and mesos, and macro, with thinner lines showing all individual channels.(D.) Time-frequency power spectrogram, from top to bottom, of an example micro, meso, and macro electrode for the recorded HFO.

Table 1 .
Cortical stimulation from multi-scale array.