{"paper_id":"860ab524-6b16-4799-8f63-c597f9ea254f","body_text":"Explant analysis of Utah electrode arrays implanted in human cortex for brain-computer-\ninterfaces \n \nKevin Woeppel 1,2, Christopher Hughes 1,2,3, Angelica J. Herrera 1,2,3, James Eles 1, Elizabeth C. \nTyler-Kabara2,5, Robert A. Gaunt1,2,3,4, Jennifer L. Collinger1,2,3,4, Xinyan Tracy Cui1,2 \n \n1 Department of Bioengineering, University of Pittsburgh, Pittsburgh, PA.  \n2 Center for the Neural Basis of Cognition, Pittsburgh, PA. \n3 Rehab Neural Engineering Labs, Pittsburgh, PA \n4 Department of Physical Medicine and Rehabilitation, University of Pittsburgh, Pittsburgh, PA. \n5 Department of Neurosurgery, The University of Texas at Austin, Austin, TX \n \nAbstract \nBrain-computer interfaces are being developed to restore movement for people living with \nparalysis due to injury or dis ease. Although the therapeutic potential is great, long-term stability \nof the interface is critical for widespread clinical implementation. While many factors can affect \nrecording and stimulation performance includ ing electrode material stability and host tissue \nreaction, these factors have not been investigated in human implants. In this clinical study, we \nsought to characterize the material integrity and biological tissue encapsulation via explant \nanalysis in an effort to identify factors that influence electrophysiological performance. \nWe examined a total of six Utah arrays explanted from two human participants involved in \nintracortical BCI studies. Two Pt arrays were implanted for 980 days in one participant (P1) and \ntwo Pt and two iridium oxide (IrOx) arrays were implanted for 182 days in the second participant \n(P2). We observed that the recording quality followed a similar trend in all 6 arrays with an initial \nincrease in peak-to-peak voltage during the first 30-40 days and gradual decline thereafter in \nP1. \nUsing optical and two-photon microscopy (TPM) we observed a higher degree of tissue \nencapsulation on both arrays implanted for longer durations in participant P1. We then used \nscanning electron microscopy and energy dispersive X-ray spectroscopy to assess material \ndegradation. All measures of material degradation for the Pt arrays were found to be more \nprominent in the participant with a longer implantation time.  Two IrOx arrays were subjected to \nbrief survey stimulations, and one of these arrays showed loss of iridium from majority of the \nstimulated sites. Recording performance appeared to be unaffected by this loss of iridium, \nsuggesting that the adhesion of IrOx coating may have been compromised by the stimulation, \nbut the metal layer did not detach until or after array removal.  \nIn summary, both tissue encapsulation and material degradation were more pronounced in the \narrays that were implanted for a longer duration. Additionally, these arrays also had lower signal \namplitude and impedance. New biomaterial strategies that minimize fibrotic encapsulation and \nenhance material stability should be developed to achieve high quality recording and stimulation \nfor longer implantation periods.   \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \nNOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.\n\nIntroduction \nIntracortical brain-computer interfaces (BCIs) can restore function for people affected by \nsignificant paralysis by allowing the user to control an effector or assistive device with signals \nrecorded in the brain. In recent years intracortical implants in motor cortex have been used for \nBCI control in primates and human participants with up to 10 degrees of freedom (Ajiboye et al. \n2017; Bouton et al. 2016; Collinger et al. 2013; Hochberg et al. 2006; Santhanam et al. 2006; \nVelliste et al. 2008; Wodlinger et al. 2014). More recently, somatosensory feedback has also \nbeen added to these systems by stimulating through electrodes in the somatosensory cortex \n(Armenta Salas et al. 2018; Fifer et al. 2020; Flesher et al. 2016; Flesher et al. 2019; Flesher et \nal. 2021; Hughes et al. 2020; Hughes et al. 2020). Given that intracortical BCIs require surgical \nimplantation, they must be stable over many years to be clinically viable. This issue has been \nstudied in both humans and primates, demonstrating that signals can be reliably recorded from \nelectrodes in the motor cortex for over 6 years when devices do not fail, although there is \nconsiderable inter-subject variability and signals typically deteriorate over time (Bullard et al. \n2020; Chestek et al. 2011; Downey et al. 2018; Hughes et al. 2020; James et al. 2013; Simeral \net al. 2011; Suner et al. 2005).  \nChanges in recorded activity can be caused by many factors including movements of the \nelectrodes relative to the brain, encapsulation of the electrode sites, as well as material \ndegradation and failure (Kozai et al. 2015; Prasad et al. 2014; Woeppel et al. 2017). These \nfactors can be broadly grouped into multiple failure categories, including material and biological \nfailure (James et al. 2013). \nBiological failures can occur as a result of the host tissue reactions to the implant. The traumatic \nnature of the implant leads to glial activation and encapsulation of the implant in a glial sheath \n(Polikov et al. 2005; Salatino et al. 2017). The glial sheath creates a physical barrier between \nthe electrode and the neurons, while the extensive inflammation damages healthy neurons and \nmay cause a neuron dead zone around the implant (Buzsáki 2004; Schwartz et al. 2006). One \nrecent study examining brain tissue from a human patient implanted with a Utah microelectrode \narray for seven months found a substantial degree of tissue damage which correlated with \ndecreased recording performance.(Szymanski et al. 2021) In addition to central nervous system \n(CNS) reactions, the meninges can grow under the electrode. Meningeal encapsulation is highly \ncollagenous and originates from non-CNS tissues. Substantial undergrowth of meningeal \ntissues can result in displacement of the electrode sites or complete ejection of the device from \nthe CNS. (Woolley et al. 2013) Subsequent device ejection is the most prevalent cause of \nchronic device failure in non-human primates, accounting for nearly 30% of chronic failure \n(Barrese et al. 2016; Dunlap et al. 2020) Longer experimental times increase the chance of \nmeningeal undergrowth and eventual ejection of the recording device from the host tissues \n(Barrese et al. 2016; Degenhart et al. 2016; Rousche and Normann 1998). \nMaterial failures include metal corrosion, insulation cracking, and insulation delamination. These \nmaterial failure modes often increase in likelihood as time progresses. The parylene-C \ninsulation commonly used for Utah style intracortical arrays can crack and delaminate, shunting \ncurrent to the biological tissues (Caldwell et al. 2020; Prasad et al. 2014; Schmidt et al. 1988; \nXie et al. 2014). The metal tips of Utah arrays, most commonly platinum or iridium oxide, are \ngenerally stable in vitro, but may be eroded away by aggressive stimulation (Negi et al. 2010) or \nthe comparatively harsh in vivo  environment (Negi et al. 2010). Furthermore, use of the \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\nelectrodes for stimulation can impact the rate of tip degradation (Cogan 2008; Gilgunn et al. \n2013).  \nTo establish stimulation limits for these clinical studies, experiments were performed in non-\nhuman primates and showed that frequent microstimulation over six months did not cause more \nloss of neurons around the electrode tips than insertion of the devices themselves and that \nstimulation had no behavioral effe ct for tasks that r equired tactile feedback (Chen et al. 2014; \nKim et al. 2015). Using these established parameters, we would not expect stimulation to cause \nfurther damage to the brain tissue after implantation or have deleterious effects on behavior. In \nfact, stimulation over five years in a participant with these established parameters has not \nresulted in significant differences in signal between stimulated and non-stimulated arrays and \ndetection thresholds have improved over time (Hughes et al. 2020). However, to our knowledge \nthere have been no post-implant examinations of the material properties of intracortical arrays \nimplanted in humans. Here we examine the extent to which any material degradation occurred \non explanted human intracortical electrodes, which will aid in the design and development of \nrobust BCIs for long-term clinical use. \nIn this work, electrodes explanted from two human participants were examined to determine the \nextent of tissue encapsulation and material failure and to assess how these factors affected \nchronic recording performance. These electrodes were implanted for different lengths of time \nand were surgically explanted: 987 days for the two arrays in participant 1 (P1) and 182 days for \nthe four arrays in participant 2 (P2). Both arrays in P1 and two of the arrays in the P2 had \nplatinum tips and were used for recording only, while the other two of the arrays in P2 had \nsputtered iridium oxide (IrOx) tips and were used for both stimulating and recording (Negi et al. \n2010). First, the extent and nature of the tissue encapsulation of the arrays was investigated \nusing optical microscopy and two-photon microscopy (TPM). Following this, the electrode arrays \nwere examined with scanning electron microscopy (SEM) and energy-dispersive x-ray \nspectroscopy (EDS) to evaluate the extent of material damage. Finally, we compared the results \nof these analyses to endpoint recording performance of the devices and characterized the \nrelationship between electrical\n stimulation and material degradation.  \n1. Methods \n1.1 Participants \nThese studies (NCT01894802 and NCT01364480) were conducted under Investigational \nDevice Exemptions from the U.S. Food and Drug administration and were approved by the \nInstitutional Review Boards at the University of Pittsburgh (Pittsburgh, PA) and the Space and \nNaval Warfare Systems Center Pacific (San Diego, CA). Informed consent was obtained before \nany study procedures were conducted. Two participants were implanted with microelectrode \narrays in the brain. The first subject (P1) was implanted with two intracortical Pt microelectrode \narrays (4 mm × 4 mm, Blackrock Microsystems, Salt Lake City, UT, USA) each with 96 wired \nelectrode shanks (length 1.5 mm) in a 10x10 grid in the participant’s left motor cortex (Figure 1). \nThe second subject (P2) was implanted with two Pt microelectrode arrays (Blackrock \nMicrosystems, Salt Lake City, UT) in the left somatosensory cortex and two iridium oxide (IrOx) \nmicroelectrode arrays in the left posterior parietal cortex. Each Pt array in the somatosensory \ncortex consisted of 88 wired electrodes in a 10x10 grid while each IrOx array in the posterior \nparietal cortex consisted of 32 wired electrodes distributed throughout a 6x10 grid (Figure 1). \nFollowing implantation of the arrays into P2, it was discovered that the implant locations were \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\nposterior to the intended sites. Following which, the pedestals were removed, and a second \nimplantation was performed two months later. \n1.2 Neural recording and signal quality metrics \nNeural data were collected for both P1 and P2 using Neuroport Neural Signal Processors \n(Blackrock Microsystems, Salt Lake City, UT). At the beginning of each test session, a threshold \nfor all channels was set at -5.25 (P1 before day 565) and -4.5 (all other test sessions) times the \nroot-mean-square voltage. Data were collected across 287 sessions spanning 33 months for P1 \nand 40 sessions across four months for P2. No recordings were done for the final two months of \nP2’s implant as the percutaneous pedestal connectors had been removed to prepare for the \nreimplant. \nOne of the main goals of the clinical study was to provide the participants with high degree-of-\nfreedom control of a robotic arm. To accomplish this, participants performed a brain-computer \ninterface calibration paradigm at the beginning of a test session. We used three minutes of data \ncollected during this calibration procedure to run spike sorting analyses offline. The sorting \nmethod, described in detail in Downey et al., 2018 (Downey et al. 2018) used principal \ncomponent analysis (PCA) to separate units, defined as threshold crossings from an individual \nelectrode, based on the similarity of their waveform shape. Characteristics for each unit were \nthen calculated. Peak-to-peak voltage (Vpp) was defined as the voltage difference between the \npeak and the trough of the average waveform for each unit. Since there could be more than one \nunit identified per electrode, the unit with the maximum Vpp was chosen to represent the signal \nquality for the given electrode. Electrodes were considered to be viable if they contained \nwaveforms with a minimum Vpp of 30 µV and a minimum firing rate of 0.25 Hz.  \n1.2.1 Impedances\n \nElectrode impedances were measured for both participants using the NeuroPort patient cable \ndata acquisition system (Blackrock Microsystems, Salt Lake City, UT). For P1, impedances \nwere measured at the beginning of a test session once a month. Impedances values for P2 \nwere measured at the beginning of each test session. The system delivered a 1 kHz, 10 nA \npeak-to-peak sinusoidal current to each implanted electrode for one second. \n1.2.2 Intracortical stimulation and calculated metrics\n \nSeven test sessions across approximately one month involved microstimulation on the IrOx \narrays. Stimulation was delivered using a CereStim R96 multichannel microstimulation system \n(Blackrock Microsystems, Salt Lake City, UT). Pulse trains consisted of cathodal phase first, \ncurrent-controlled, charge-balanced pulses delivered at frequencies from 20-300 Hz and at \namplitudes from 1-100 \nμ A. The cathodal phase was 200 μ s long, the anodal phase was 400 μ s \nlong, and the anodal phase was set to half the amplitude of the cathodal phase. The phases \nwere separated by a 100 \nμ s interphase period. Stimulus pulse trains were varied in terms of \namplitude, frequency, and train duration.  \nThe voltage transients associated with each stimulus pulse were recorded using National \nInstruments data acquisition modules. Voltage traces were displayed in real time using LabView \nand saved to disk for analysis. Interphase voltage was measured as the voltage at the end of \nthe interphase period immediately prior to the anodal phase for a given stimulation pulse. The \ntotal charge delivered to each electrode was calculated across all stimulation experiments using \nthe charge delivered during the cathodal phase.  \n1.3 Explanted array handling before imaging\n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\nThe two Pt arrays in P1 were explanted on day 987 post-implant and the four arrays in P2 were \nexplanted on day 182. Following explantation, all arrays were removed from their wire bundles \nby clipping the wires proximal to the probe and were washed with saline. The P1 arrays were \nimmediately fixed in formalin and then transferred to PBS bath for storage. \nImmunohistochemical staining procedure was performed on these two arrays with the goal of \nidentifying neuron (NeuN) and microglia/macrophage (Iba-1). The staining process involves \nincubation of the arrays with primary antibodies solutions overnight, with secondary antibodies \nfor 4 hours followed by Hoescht solution for 20 min for nuclei staining. The antibody staining \nwas unsuccessful, and only nuclei stain was used for the tissue analysis. The P2 arrays were \nfixed 2 months post-implant, and one of the Pt arrays had visible tissue encapsulation and was \nimaged using TPM. Because these arrays were not immediately fixed, we did not perform \nimmunostaining, and only characterized the collagen structure, which can be stable without the \nfixation.  \nAfter optical and TPM imaging, two arrays explanted from P1 were sent to the FDA for initial \nanalysis. The arrays were initially imaged with an environmental SEM, then enzymatically \ncleaned with Asepti-Zyme neutral pH enzymatic instrument presoak/cleaner (4ml in 250ml \nsaline) at 37°C for 90 minutes, followed by Getinge Clean Enzymatic detergent (1ml in 250ml \nsaline) at 37°C for 90 minutes, and then by MetriZyme detergent (1ml in 250ml saline) 37°C for \n90 minutes. Samples were then thoroughly washed with water and air dried, ready for SEM \nimaging. This process was effective at removing some of the tissue and revealing the electrode \ntip/shank for material analysis. Arrays from P2 did not undergo the enzymatic cleaning \nprocedure. All arrays were stored adhered to copper tape, tips up.  \n1.4 Electrode Imaging \n \nExplanted electrodes were first characterized by optical and two-photon microscopy to assess \nthe degree of tissue encapsulation. For TPM, we used a two-photon laser scanning microscope \nwith a Bruker scan head (Prairie Technologies, Madison, WI), TI:sapphire laser tuned to 920 nm \n(Mai Tai DS; Spectra-Physics, Menlo Park, CA), light collection through non-descanned \nphotomultiplier tubes (Hamamatsu Photonics KK, Hamamatsu, Shizuoka, Japan), and a 10x or \n16x, 0.8 numerical aperture wate r immersion objective (Nikon In c., Milville, NY). Laser power \nwas maintained between 20-40 mW. For each electrode tip, Z-stacks were collected with filters \nto resolve second harmonic generation (SHG) at half the laser wavelength (~460nm), which \nenabled intrinsic imaging of collagen-I representing the meningeal encapsulation. Images along \nthe length of the electrode shanks were collected as Z-stacks. Z-stack images were either \ncollected at specific regions of interest, or in a grid at all locations across the face of the \nelectrode array. Grid images were automated by the Prairie software with a 10% overlap \nbetween images. All image stitching and subsequent image processing was conducted with \nImageJ software (NIH).  Electrode integrity was characterized by scanning electron microscopy \n(SEM) and energy-dispersive x-ray spectroscopy (EDS). Samples were washed, dried under \nalcohol, and sputter-coated with 4nm Au/Pd. Images were taken by JSM 6335F electron \nmicroscope. EDS was taken by Zeiss Sigma 500VP, excluding Au and Pd from quantification.  \nUsing the SEM and optical images, a qualitative category of ‘non-degraded/unencapsulated’ or \n‘degraded/encapsulated’ was assigned to each electrode based on the degree of damage to the \ntip or shank, or the level of encapsulation around the electrode ( Figure S1). Arrays explanted \nfrom P1 were more extensively cleaned prior to imaging, and the encapsulation score was \nbased on optical images of the explanted arrays. Encapsulation on arrays from P2 was \ndetermined by examining the SEM images. Degraded electrode tips were defined as having \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\nobvious and substantial surface defects in the metal coating, including pitting of the metal, \nflaking of the metal, and ex posure of the underlyi ng silicon. Degraded shanks were defined \nrelative to the parylene insulation, with defects including insulation cracking along the shank, \npeeling of the insulation away from the shank near the tip, and other obvious defects in or below \nthe insulation. These categories were compared to EDS images, confirming the \npresence/absence of metalation at the tip (Pt/IrOx). Electrodes which could not be quantified, \ndue to breakage during removal or gross encapsulation, were assigned a null score and \nexcluded from analysis. \n1.5 Statistics\n \nChanges in signal and impedances over time were assessed using linear regression. For \nimpedances, data were log-transformed because data did not follow a linear trend. For P2 \nimpedances and Vpp, we excluded data prior to day 30 for regression because the impedances \nmeasured in this range were highly variable.  \nTotal charge delivered, minimum interphase voltages, and charge delivered after exceeding an \ninterphase voltage of -0.6V were compared between the two electrode arrays that had received \nstimulation using Mann-Whitney tests. We used a non-parametric test because the data was \ndetermined to not be normally distributed using an Anderson-Darling test. We used a Fisher \nexact test to determine if there was a significant relationship between an electrode’s material \nproperties (undamaged or damaged) and the length of implantation (Pt arrays in P1 vs. P2) or if \nit received stimulation (P2 IrOx arrays, yes or no). We further quantified if there was a \nrelationship between both total charge injected and charge injected with interphase voltages \nbelow -0.6V on stimulated electrodes and their material properties (undamaged or damaged) \nusing logistic regression. Electrode categories were compared to impedances and Vpp using \nMann-Whitney tests. We used a non-parametric test because the variances between groups \nwere not the same.  \n2. Results \n2.1 Signal amplitude and impedances decreased over time\n \nChanges in the impedances and peak-to-peak voltages over time were observed on implanted \nelectrodes in both participants (Figure 2). Impedances decreased over time on electrodes \nimplanted in P1 (p<0.001, log-transformed linear regression, Figure 2A). For P2, the starting \nimpedances of  IrOx electrodes were lower than the platinum electrodes, which is consistent \nwith the manufacturer’s specification (Negi et al. 2010). From day 1 to 20 we observed an \nincrease in impedances. The initial increase in impedance reversed after one month (30 days), \nand a significant downward trend in impedances was observed until the end of recording for \nboth the IrOx (p<0.001, linear regression) and platinum arrays (p<0.001, linear regression). \nImpedances gathered from P1 eventually stabilized after approximately two years. The \ndifference between the final impedance values recorded in P1 and P2 can be explained by the \ndifference in length of implantation. Previous studies have determined that impedance values of \nstimulated and non-stimulated intracortical electrodes decrease dramatically over the first \ncouple of years after implantation in humans (Hughes et al. 2020) and monkeys (Chestek et al. \n2011; Suner et al. 2005). Since P2 was implanted for a significantly shorter period, we would \nexpect the electrode impedance values to be larger and more variable, which the data supports.  \nIn the same manner as the impedance measurements, an initial increase in Vpp was observed \nfor both P1 and P2. However, after an increase in the first 30 days, the measured Vpp from P1 \nand P2 exhibited a downward trend (p<0.001, linear regression, Figure 2C,D). The rates of \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\ndecrease in the Vpp between day 30 and 120 for P1 and P2 were -4.0 µV/month and -4.86 \nµV/month, respectively. Median Vpp decreased by 52% across 550 days in P1 and by 14% \nacross 90 days in P2. The median Vpp for P1 leveled off at approximately 25-30 µV. \n2.2 Encapsulating tissues were apparent on multiple arrays \nBased on the gross optical micrographs, both P1 arrays had a significant degree of adherent \ntissue on the electrode base and shanks. For the P2 arrays, one of the Pt arrays and one of the \nIrOx arrays showed some tissue deposits while the other arrays were clean. The nature of the \nencapsulating tissue was examined with TPM, measuring the second harmonic signal \ncharacteristic of collagen. For the more heavily encapsulated P1 arrays, the encapsulation \nsheet was found both along the shanks of the array (Figure 3A-D) and at the base (Figure 3E-H \n(P1)). Strong second-harmonic signal within the tissue sheet confirmed that it was primarily \ncomposed of collagen-I fibers (Figure 3C,D,G,H). After further examining the indicated \nelectrodes and staining for cell nuclei, we observed that the encapsulation was highly \ncellularized (Figure 3C,D,G,H). In addition, the encapsulation continued down the shank of the \nelectrode, with cellular and collagenous material detected along the shanks and tips of the \narray. On the underside of the array at the base of the shanks the encapsulation was not \nhomogenous, instead exhibiting greater second harmonic signals nearer to the edges (Figure \n3G) while having greater cell density nearer the center (Figure 3H). SHG imaging is also a good \ntool for detecting blood vessels because of the strong presence of collagen in the vessel wall, \nhowever we did not observe clear blood vessel structure in the P1 explants. \nFor the posterior Pt array in P2, the encapsulation tissue covers the whole array (Figure 4B) and \nthe TPM imaging from the side revealed significant tissue covering the majority of the electrode \ntips.  Here, we identified clear vascular architecture in the encapsulation tissue (Figure 4D, \nhighlighted in blue). The blood vessel in the encapsulation tissue was traced and super-imposed \nto the image of pial vasculature observed pre-implantation (Figure 4E). As can be seen in \nFigure 4F, the blood vessel traces match the pia vasculature. This indicates that the blood \nvessels observed to be at the tip of this array were pial blood vessels. Two mechanisms may \nlead to this: 1) the array did not fully penetrate the pia at the time of implantation; 2) the array \nwas successfully implanted in the brain parenchyma and the pia membrane was pulled out with \nthe array. Since we were able to obtain high quality single unit recordings from this array even \nfrom the affected region, the first potential mechanism was ruled out. Therefore, we conclude \nthat at least some of the tissue on this explant is pia membrane that was pulled out with the \ndevice, not fibrotic scar tissue formed as the result of foreign body reaction.  \n2.3 Length of implantation impacts the degree of material degradation and fibrous \nencapsulation.  \nBased on the optical, TPM and SEM images, electrodes were assigned a binary score for the \ntip, shank, and degree of fibrous encapsulation (Figure 5). Electrodes that appeared to be \nbroken or damaged by implantation/explantation were excluded from analysis. Tips and shanks \nwere evaluated separately to examine the effects of both tip metallization and electrode \ninsulation on device performance. The number of electrodes for each group are displayed in \nTable 1, excluding electrodes which were not wired or used for recording or stimulation. \nDifferences in the total number of electrodes receiving a tip category (n=387), shank category \n(n=413), and encapsulation (n=380) are due to damage to the electrodes or excess \nencapsulation preventing the assignment of a proper category (Figures S1-3).  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\nCategories assigned to P1 and P2 platinum arrays were compared to identify any potential \nchanges in material deterioration or encapsulation which may be attributed to the length of \nimplantation (Table 2). We found that both measures of material degradation (tip and shank \ndamage) were more prominent for longer implantation times (27.8% tip damage for P1 and \n9.9% for P2, 15.2% shank damage for P1 and 1.7% for P2, p<0.001 for both). We also found \nthat the degree of encapsulation is more significant for longer implants with 72.4% for the P1 \narrays and 49.4% for the P2 arrays (p<0.001).  \n2.4 Stimulation resulted in electrode damage on one stimulating array but not the other\n \nTwo IrOx arrays implanted in P2 received a low amount of total charge (<160 µC per electrode \nsite). Each of the two stimulated IrOx arrays had 60 electrodes, half of which electrically \nconnected and used for stimulation. Preimplant optical images of the arrays did not show any \nvariation between arrays. The stimulated electrode sites are arranged primarily in a \ncheckerboard fashion. SEM shows that the lateral array had a high degree of tip and shank \ndegradation (Figure 6A).  Interestingly, tips and shanks showing visible damage appeared to \ncoincide with the electrodes that were used for stimulation. Furthermore, EDS revealed that \nstimulated tips had lower iridium content than non-stimulated tips (Figure 6B). The loss of \nmetallization for the lateral stimulating a rray occurred only on th e electrodes  used for \nstimulation. The medial array did not show this pattern (Figure 6C). The damage scores for \neach electrode tip and shank are summarized in the Figure 6D,E. The checkerboard pattern of \ndamages of the lateral array is clearly seen, which correspond very well with arrangement of the \nstimulation electrodes. The medial stimulating array has overall much less observable material \ndamage but more tissue encapsulation. Of the 62 electrodes used for stimulation on both \narrays, 56 were analyzed, of which 23 had notable tip degradation, 21 of which were located on \nthe lateral electrode array. Metal loss, and the corresponding decrease in iridium signal, was not \nobserved on any non-stimulated electrodes. These results are summarized in Table 3.   \nDelivered charge and measured interphase voltages were compared to the material \ndegradation. The amount of stimulation provided was quantified by both the total charge \ndelivered and number of pulses delivered. Although the mean amount of charge injected on the \nlateral array was greater, it was not significantly different than the mean charge injected on the \nmedial array (Mann-Whitney test, p = 0.22). The medial array contained the three electrodes \nwith the most charge delivered, none of which displayed observable material degradation. \nHowever, we examined the minimum voltage during the interphase period (Figure 6F) and \nfound that the lateral array electrodes experienced higher voltage excursions on average than \nthe medial array electrodes (mean minimum voltage was = -1.7 V for the lateral, and  -1.1 V for \nthe medial array). Furthermore, there was a significant relationship between the total charge \ninjected at voltages more negative than -0.6V (Figure 6G) and the tip score (p = 0.025, crit-p = \n0.034, logistic regression) and shank category (p = 0.023, crit-p = 0.034, logistic regression) on \nthe lateral stimulating array. There was no relationship between total charge injected at voltages \nless than -0.6V and tip category (p = 0.60, logistic regression) or shank category (p = 1, logistic \nregression) on the medial array.  \nWe found no significant differences in recording quality (Vpp) (Figure 6H) between the damaged \nand non-damaged electrodes on the two stimulation arrays, after excluding the encapsulated \nelectrodes.  \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\n3. Discussion  \nIn order for BCIs to become a viable therapy, the longevity of the devices and mechanisms of \nfailure must be well understood. Effective electrode design requires knowledge of the stability of \nthe materials in the harsh in vivo environment and the effects of gradually accumulating damage \nto the device. However, the relationships between chronic material degradation and device \nperformance are poorly understood. The effects of material degradation on performance in \nhuman subjects is further complicated by the limited number of human subjects and the even \nsmaller amount of explanted human BCI arrays. In this work, the in vivo performance of human \nneural electrode arrays was compared to the material integrity after explant. We have found \nsigns of material degradation on all electrode arrays, with longer implantation times correlating \nwith an increased number of degraded electrodes (Table 2). Additionally, biological tissue \nencapsulation on the explanted device was also documented as another potential factor to \ninfluence recording quality. The biological encapsulation tissues were highly collagenous and \nalso highly cellularized, and appear to form in a time dependent manner, increasing with the \nlength of implantation. A similar form of tissue response has been observed in a post mortem \nanalysis of tissue surrounding a MEA implanted for seven months.(Szymanski et al. 2021) \nFurther, the nature of the encapsulation at the periphery of the array and the center is different. \nTogether, these results suggest that the encapsulation originated from the meninges, as \nopposed to the CNS.  \n3.1 Encapsulation and material degradation were both related to the length of implantation \n \nMaterial and biological failure modes are most common on longer time scales (James et al. \n2013), and it was expected that material degradation and collagenous encapsulation would \nincrease with longer implant times. Indeed, we observed that the arrays implanted in P1 \nexhibited greater degrees of material degradation and encapsulation than those in P2 which \nwere implanted for a much shorter length. We also observed a characteristic decline in \nrecording performance and impedances with longer implantation times.  \nImpedance measurements are often used to determine the integrity of electrodes, while also \nserving as a method of investigating the interface between the electrode and the host tissues \n(Lago et al. 2016; Thakore et al. 2012). Previously, we have reported that in rats, complete \nfibrous encapsulation of the electrode resulted in lower 1kHz impedance compared to partial \nencapsulation (Cody et al. 2018). Complex impedance spectra analysis performed in the \naforementioned study revealed unique features in the Nyquist plot that corresponds to an \nextracellular resistance component, which is smaller in the fully encapsulated device than the \npartially encapsulated device. This may be counterintuitive initially, but can be explained by a \nfew mechanisms. First, the composition of the encapsulation tissue is high in collagen and less \nresistive than highly cellular and myelin rich brain tissue. Secondly, if the fibrotic growth at the \nbase pushes the array up, a liquid filled gap will be formed between cone shaped shanks and \nthe tract, creating a less resistive current path. Due to the limitation in our instrumentation in this \nstudy, impedance data were only obtained at 1 kHz preventing us from measuring complex \nimpedances, but it is plausible that a similar effect may have occurred here. Full spectrum \nimpedance recording in future studies could dissect the contributions from tissue encapsulation \nand material changes and determine whether the same factors are relevant here. However, \nsuch measurement needs to meet the regulatory requirements associated with clinical studies. \nThe decreases in impedance over time could also be a result of degradation of the electrode \ntips and shank insulation which leads to increased electrochemical surface area. Interestingly, \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\nwe found no relationships between the impedance of the electrode at 1kHz and the Vpp during \nrecording for Pt arrays (Figure S4). This is not surprising as impedance is only a measure of the \nelectrochemical properties of the electrode and the electrode/tissue interface and does not \naccount for biological variables such as distance from the electrode to the neuron or health of \nthe host tissues, which are more relevant to Vpp. Impedance has previously been shown to be \nan unreliable predictor of recording performance in rodents and non-human primates (Cody et \nal. 2018; Jiang et al. 2014). Another important material factor that likely contribute to the uniform \nreduction of impedance on all arrays is the silicone hermetic sealing failure above the arrays \nfrom the wire bundle to the pedestal, which should be characterized in future studies. \nThe observed collagenous encapsulation of the arrays has been observed in rodent and non-\nhuman primate studies (Cody et al. 2018; Degenhart et al. 2016; James et al. 2013). \nEncapsulation of the electrode tip region can isolate the electrode from nearby neurons, \nresulting in a lowered Vpp. In addition, the collagenous material grown at the base of the array \nplatform can lift the electrode up and away from the original target neurons, also resulting in Vpp \ndecrease (Cody et al. 2018; Degenhart et al. 2016).  Both tissue growth at the tips and the base \nhave been observed from the explanted devices which may contribute to the degradation of Vpp \nover time in human subjects.  \n3.2 Stimulation at more negative voltages may drive material damage under certain \ncircumstances \nThe stimulation parameters used in this study were based on studies from non-human primates \nshowing that these parameters had no additional effects on cortical tissue apart from implanting \nthe devices themselves, had no behavioral effect on the animal, and had limited effects on the \nelectrode tissue interface (Chen et al. 2014; Kim et al. 2015).  Here, we found that electrical \nstimulation induced damage on one of the two devices. On the lateral array, de-metallization \nwas visible under SEM and detected by EDS only on the stimulated electrodes, indicating that \nstimulation was the cause of the metal loss. The reason that stimulation caused material \ndamage on the lateral array but not the medial array is unclear. One notable difference between \nthe lateral and medial array is that the medial had higher degree of tissue encapsulation. This \ncan be a result of a higher degree of vascular damage or less stable fixation in vivo. While the \nspecific reason for this cannot be determined, impedances were lower on the medial array \n(Figure S5). The decrease in impedance then could have resulted in lower amplitude voltage \nexcursions during stimulation, decreasing the likelihood of material damage. Indeed, we found \nthat the lateral array had more negative interphase voltages (mean = -1.7V) when compared to \nthe medial array (mean = -1.1V).  \nMore material damage was found on the lateral stimulating array in P2 which experienced \nhigher cathodic interphase voltage amplitude. The interphase voltage is analogous to the \nmaximum cathodic electrode potential (E\nmc) measured during charge injection limit (CIL) \nexperiments performed in vitro. Emc with values more negative than -0.6V (vs Ag/AgCl) are often \nconsidered to be unsafe due to irreversible water hydrolysis occurring at the electrode which \ncould cause hydrogen gas production and pH increases (Cogan et al. 2005). Such reactions \ncould lead to delamination of the IrOx coating even with a small number of pulses. We do not \nexpect the median interphase voltage to be directly comparable to in vitro CIL measurements \ndue to the two-electrode setup and increased variables introduced from the biological \nenvironment, but we expect the general relationship between voltage and interfacial reactions to \nhold. Besides the fact that lateral array experienced higher voltage excursion on average, we \nfound a significant correlation between the charge injected below -0.6V and damage on the \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\nlateral stimulating array. These results indicate that stimulation, beyond a certain voltage \nthreshold, may damage electrodes in a dose (charge) dependent manner.  \nPartial and complete loss of SIROF from Utah arrays upon continuous stimulation has been \nreported in previous in vitro studies (Negi et al. 2010), but the stimulation doses in these \nprevious studies applied were much higher (7 h of continuous stimulation above 60nC). One \npotential explanation is that the stimulation on the lateral array only weakened the adhesion of \nthe IrOx coating and the coating was stripped from the electrode during or after explantation. \nAlternatively, variations could also be a result of batch-to-batch difference in fabrication where \nthe lateral array received poorly adhered IrOx coating. Notably, the electrodes damaged by \nstimulation performed just as well in recording as the undamaged electrodes. This surprising \nresult indicates that despite the IrOx delamination and insulation cracking, the electrode is \ncapable of recording neural signals (Hughes et al. 2020).  \nThe biocompatibility of IrOx coatings has been widely studied and validated for stimulation and \nrecording applications (Cogan 2008; Cogan et al. 2005; Hughes et al. 2020; Lee et al. 2002; \nNegi et al. 2010; Negi et al. 2010). In another of our studies, stimulating electrodes for over five \nyears in a human participant did not result in worse signal recordings when compared to \nrecording electrodes(Hughes et  al. 2020). Furthermore, the ab ility to evoke sensation on \nstimulated electrodes only improved over time. Based on our observations here, this could be \nbecause a) the material damage caused by stimulation is idiosyncratic, and stimulation didn’t \nresult in damage on the arrays of this 5-year study or b) material damage caused by stimulation \nhad no effect on the electrode’s ability to record or stimulate. Discerning between the two is \ndifficult, as studying the in vivo  properties of the electrodes in parallel with the material \nproperties is not possible in humans. Analysis will need to be conducted on these arrays that \nreceived much higher levels of stimulation after explant. Additionally, further animal studies \nusing the stimulus parameters used in our study and assessing damage and changes in \nrecording over time could provide insight here.  \n3.3 Implications for future intracortical electrode arrays\n \nOverall, our results show that both material integrity and recording performance of human \nintracortical electrodes decrease over time. Degradation was observed on both electrode tip and \nthe shank insulations. We have also observed different degree of tissue encapsulation both at \nthe array base, middle of shank and tips of the arrays. Since we do not have real time data of \nthese material and tissue changes, and explant analysis only provides a partial picture at the \nend point, we cannot accurately correlate material and biological factors to recording outcome. \nMultiple human studies have demonstrated that intracortical electrode recordings can enable \nbrain-computer interface control of computer cursor and robotic arms for years after \nimplant,(Bullard et al. 2020) yet the observations in the current study support the need for \nstrategies for increasing material durability and decreasing fibrous encapsulation in order to \nfurther improve human BCI recording quality and longevity. Additionally, on one implanted array, \nwe observed clear iridium loss as a result of stimulation, which correlated to more charge \ninjected at more negative voltages. Further research on improving metal adhesion and \ndeveloping real time electrode potential monitoring method during stimulation will eliminate such \nincidences. \n \nAcknowledgements \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\nWe thank the participants for their dedication to this study. The recording and stimulation study \nwas partly funded by the Defense Advanced Research Projects Agency’s (Arlington, VA, USA) \nRevolutionizing Prosthetics program (contract number N66001-10-C-4056). The views \nexpressed herein are those of the authors and do not represent the official policy or position of \nthe Department of Defense or US Government. The explant materials and tissue analysis were \nsupported by the National Institute of Health grant R01NS110564, R01NS062109 and \nR01NS089688. We thank Dr. Pavel Takmakov of the Federal Drug and Agriculture for the initial \ndiscussion and cleaning protocols for SEM.\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\nReferences \nAjiboye, A. B ., Will et t, F. R. , Young, D. R., Memberg, W . D., Murp hy, B. A ., Mille r, J .  P., et al. (2017). \nRestor atio n of reaching and grasping mo vements thr ough brain-cont rolle d muscle stimulati on in a \nperson with t et raplegia : a proof-of-concept demons tra tion . The Lan cet  389, 1821 -1830. \ndoi:ht tps://doi.o rg/10.1016/S0140-6736 (17)30601-3  \nArment a Salas, M ., Bashfo rd, L., Kellis, S .,  Jafari, M ., Jo , H., Krame r, D., e t al. (2018). Proprioceptive a nd \ncutaneo us sensatio ns in humans elicit ed by intracor tical microstimula tion . eLife 7, e32904. \ndoi:10.7554/eLife.32904  \nBarr ese, J . C., Ace ros, J . and Donoghue, J.  P. (2016). Scanning electron microscopy of chronically \nimplanted in traco rtical micro elect rode a rrays in non-human primat es. J our nal of Neural En gine ering  13, \n026003. doi:10.1088/1741-2560/13/2/0 26003 \nBouton , C. E., Shaikhouni , A., Anne tta , N.  V., Bockbrade r, M. A ., Fri edenb erg, D. A . , Nielson, D. M ., e t al. \n(2016). Restoring cortical con trol of funct ional movement in a human wit h quadri plegia. Na t u re  533, \n247-250. doi:10.1038/natu re17435  \nBullard, A . J ., Hutchison , B. C., Le e, J ., Chestek, C. A . and Patil, P. G . (2020). Estimating Risk for Future \nIntr acranial , Fully Implant ed, Mod ular N europr osth etic Systems: A Syst ematic Re view of Hardware \nComplications in Clinical Deep Brain S timulation and E xperim enta l Human Int raco rtical Ar rays. \nNeuro mod ula tion : Tech nol ogy at t he Ne ural In terfac e  23, 411-426. \ndoi:ht tps://doi.o rg/10.1111/ner.13069\n \nBuzsáki, G. (2004). Large-scale r ecording of neuronal e nsembles. Na ture Ne urosci ence  7, 446-451. \ndoi:10.1038/nn1233  \nCaldwell, R., St ree t, M . G. , Sharma , R., Ta kmakov, P., Baker, B. an d Riet h, L. (2020) . Characte riza tion of \nParylene-C degradat ion mechanisms: In vitro reac tive accele rat ed aging model co mpared to mul tiyear in \nvivo implantation . Biom ateri als  232, 119731. doi: ht tps://doi.o rg/10.1016/j.biomaterials .2019.119731  \nChen, K. H., Dammann, J . F., B oback, J. L. ,  Tenore, F . V., Ot to, K. J., Gaun t, R. A ., e t al. (2014). The effect \nof chronic intracor tical microstimul ation on the el ectr ode –tissue in terfac e. J ourn a l of Neural Eng ineeri ng  \n11, 026004. doi:10.1088/1741-2560/11/ 2/026004  \nChestek, C. A. , Gilja, V., N uyujukian, P., Foster , J. D., Fa n, J . M., Kaufman, M . T., e t  al. (2011). Long-term \nstability of neur al prost hetic con trol sign als from silicon cortical arr ays in rhesus macaque moto r cort ex . \nJour nal of Ne ural En gineeri ng  8, 045005. doi:10.1088/1741-2560/8/4/045005  \nCody, P. A., Eles, J. R ., Lagenau r, C. F., Ko zai, T. D. Y. and Cui, X. T. (2018). Unique electr ophysiological \nand impedance signa tures b etwe en enca psulation typ es: An an alysis of biological Utah ar ray failure an d \nbenefit of a biomime tic coating in a ra t model. B i om at e r i a l s  161, 117-128. \ndoi:ht tps://doi.o rg/10.1016/j.biomateri a ls.2018.01.025\n \nCogan, S. F. (2008). Neural S timulati on a nd Recording Elect rodes . An nu al Review of Biomedi cal \nEngineeri ng  10, 275-309. doi :10.1146/an nurev.bio eng.10.061807 .160518 \nCogan, S. F., Troyk, P. R., Ehrlich , J. an d Plante, T. D. (2005). In vitro compa rison of the charge-injec tion \nlimits of activated iridi um oxide (AI ROF) and platinum-iridium micro elect rodes . IE EE Transactio ns on \nBiome dical En gin eering  52, 1612-1614. d oi:10.1109/TBME.2005.851503  \nCollinger, J. L. , Wodlinge r, B. , Downey, J. E., Wang, W ., Tyler-Kabar a, E. C., We ber ,  D. J., et al. (2013). \nHigh-performance neu ropr osthe tic cont r ol by an individual with tet rapl egia. The L ance t 381, 557-564. \ndoi:ht tps://doi.o rg/10.1016/S0140-6736 (12)61816-9\n \nDegenhart , A. D., Eles , J., Dum, R. , Misch el, J. L. , Smalianchuk, I ., Endler , B., et al . (2016). Histological \nevaluation of a chronically-implanted ele ctrocor ticographic el ectr ode grid in a no n-human primate . \nJour nal of Ne ural En gineeri ng  13, 046019 . doi:10.1088/1741-2560/13/4/046019  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\nDowney, J. E., Schwed, N ., Chase, S . M., S chwartz, A . B. and Collinge r, J . L. (2018). Intracor tical r ecording \nstability in human br ain–compu ter in terf ace users. Jour nal of Ne ural En gineeri ng  15, 046016. \ndoi:10.1088/1741-2552/aab7a0 \nDunlap, C. F., Colachis, S. C., M eyers, E. C ., Bockbrad er, M . A. a nd Fried enbe rg, D. A. (2020). Classifying  \nIntr acortica l Brain-Machi ne In terfac e Signal Disruptions Bas ed on System Perfor mance and Applicabl e \nCompensatory St rat egies: A R eview. Fro ntiers in Ne uroro bo tics 14, doi:10.3389/fnbot.20 20.558987 \nFifer, M. S. , McMulle n, D. P., Thomas, T. M., Osbo rn, L. E., Nickl, R. W. , Candre a, D. N., et al . (2020). \nIntr acortica l Microstimul ation Elicits Hu man Fingertip S ensati ons. me dRxi v 2020. 05.29.20117374. \ndoi:10.1101/2020.05.29 .20117374 \nFlesher, S . N. , Collinger, J . L., Fold es, S. T. , Weiss, J . M., Downey, J . E., Tyler-Kabar a , E. C., et al . (2016). \nIntr acortica l microstimula tion of human somatosenso ry corte x. Scie nce Tra nslati o nal Me dicin e  8, \n361ra141.  \nFlesher, S . N. , Downey, J. E., Weiss, J . M. ,  Hughes, C. L., Herr era, A. J ., Tyler-Kabar a, E. C., et al . (2019). \nRestor ed t actile sens ation imp roves neur oprosth etic arm con trol . bioR xiv  653428.  doi:10.1101/653428 \nFlesher, S . N. , Downey, J. E., Weiss, J . M. ,  Hughes, C. L., Herr era, A. J ., Tyler-Kabar a, E. C., et al . (2021). A \nbrain-compute r int erface th at evokes tac tile sensa tions improves r obotic a rm control. Scie nc e  372, 831. \ndoi:10.1126/science. abd0380  \nHochberg, L. R. , Ser ruya, M. D., F riehs, G.  M., Mukand, J. A ., Sal eh, M. , Caplan, A . H., et al. (2006). \nNeuro nal ensembl e cont rol of prosth etic  devices by a human with tetr aplegia . Na ture  442, 164-171. \ndoi:10.1038/natu re04970  \nHughes, C. L., Fleshe r, S. N ., W eiss, J. M ., Downey, J. E., Collinger, J. L. and Gaun t, R. A. (2020). Neur al \nstimulation a nd reco rding performa nce i n human somatosens ory cort ex over 1500  days. medR xiv  \n2020.01.21.20018341. doi :10.1101/2020.01.21.20018341  \nJames, C. B. , Nave en, R. , Kaivon, P., Core y, T., Carlos, V.-I., Lachlan , F., e t al. (2013 ). Failure mode an alysis \nof silicon-based intraco rtical micro elect r ode arr ays in non-human primat es. J ourn al of Neural \nEngineeri ng  10, 066014.  \nKim, S., Callier, T., Tabo t, G. A ., Ga unt, R . A., Teno re, F . V. and Bensmai a, S. J . (2015). Behavioral \nassessment of sensitivity t o intr acortic al microstimulatio n of primate soma tose nsory corte x. Proce edin gs \nof the Na tio nal A ca demy of Sciences  112,  15202. doi:10.1073/pnas.1509265112  \nKozai, T. D. Y., Catt , K., Li, X., Gugel , Z. V., Olafsson, V. T., Vazque z, A. L. , et al . (2015). Mechanical failure \nmodes of chronically implanted pl anar sil icon-based neur al prob es for laminar r ecording. B i om at e r i a l s  \n37, 25-39. doi: ht tps://doi.org/10.1016/j. biomate rials.2014 .10.040\n \nLago, N., Cest er, A ., W rachien , N., Nat ali,  M., Quiroga , S. D., Bon et ti, S., et al . (2016). A physical-based \nequivalent circui t model for an o rganic/e lectrolyt e int erface. Or gani c Elec troni cs  35, 176-185. \ndoi:ht tps://doi.o rg/10.1016/j.orgel.2016. 05.018\n \nLee, I.-S ., Whang, C.-N ., Choi, K., Choo , M .-S. and Lee, Y.-H . (2002). Characteriz atio n of iridium film as a \nstimulating neu ral el ectro de. Bi om ateri al s  23, 2375-2380. doi: https://doi. org/10.1 016/S0142-\n9612(01)00373 -8  \nNegi, S., B handa ri, R. , Riet h, L. and Solzb acher, F. (2010). In vitro compa rison of sputt ered ir idium oxid e \nand platinum-coa ted n eural implan tabl e microelect rode a rrays. Bi ome dica l Mat er ials  5, 015007. \ndoi:10.1088/1748-6041/5/1/015007  \nNegi, S., B handa ri, R. , Riet h, L., Van Wag enen, R . and Solzb acher, F . (2010). Neura l electro de degr adati on \nfrom continuous el ectrical stimul ation : Comparison of sputt er ed and activa ted iri dium oxide. Jo urnal of \nNeurosci en ce Met ho ds  186, 8-17. doi: ht t ps://doi.org/10.1016/j.jneumeth .2009.1 0.016  \nPolikov, V. S., Tresco, P. A. and R eicher t, W. M. (2005). Response of br ain tissue t o chronically implante d \nneural el ectr odes. Jo urnal of Neurosci enc e Meth ods  148, 1-18. \ndoi:ht tps://doi.o rg/10.1016/j.jneumeth. 2005.08.015  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\nPrasad, A., Xu e, Q.-S. , Dieme, R. , Sankar, V., Mayrand, R ., Nishid a, T., e t al. (2014).  Abiotic-biotic \ncharacte riza tion of Pt/Ir microel ectr ode arrays in chronic implants . Fron tiers in N euroen gin eering  7 , \ndoi:10.3389/fneng.2014.00002  \nRousche, P. J . and No rmann, R . A. (1998). Chronic recording capabili ty of the Ut ah Intr acortica l Electro de \nArray in cat sens ory cort ex. Jour nal of Ne uroscie nce Me th ods  82, 1-15. \ndoi:ht tp://dx.d oi.o rg/10.1016/S0165-02 70(98)00031 -4\n \nSalatino , J. W., Ludwig, K. A., Koz ai, T. D. Y. and Purcell, E. K. (2017). Glial respo nses to implant ed \nelectr odes in th e brain . Na ture Bi ome dic al Engin eerin g  1, 862-877. doi:10.1038/s 41551-017-0154-1  \nSanthan am, G. , Ryu, S. I ., Yu, B. M ., Afsha r, A. and Sh enoy, K. V. (2006). A high-per formance brain –\ncomputer in terfac e. Na t u re  442, 195-198. doi:10.1038/na ture04968  \nSchmidt, E. M., McInt osh, J. S . and Bak, M. J. (1988). Long-term implants of Paryl ene-C coated \nmicroelect rodes . Medi cal a nd Bi olog ical Engineeri ng a nd Co mp utin g  26, 96-101. \ndoi:10.1007/BF02441836  \nSchwartz, A . B., Cui, X. T., Webe r, Dougla s J. and Mor an, D. W . (2006). Brain-Contr olled In terfaces : \nMovement R esto ratio n with Neu ral Pros thetics . Neur on 52, 205-220. \ndoi:ht tps://doi.o rg/10.1016/j.neuron.20 06.09.019\n \nSimeral, J . D., Kim, S. P., Black, M . J. , Donoghue, J . P. and Hochberg, L. R . (2011). Neural con trol of cursor \ntraject ory and click by a human with tet r aplegia 1000 days after implan t of an intr acortical \nmicroelect rode a rray. Jour nal of Ne ural E ngine ering  8, 025027 . doi:10.1088/1741-2560/8/2/025 027  \nSuner, S ., Fellows, M . R., Vargas-I rwin, C. , Nakat a, G . K. and Donoghue, J. P. (2005). Reliability of signals \nfrom a chronically implanted , silicon-bas ed elect rode array in non-human p rimat e primary moto r cort ex. \nIEEE Transacti ons o n Neur al Systems a nd  Rehabili ta tio n Engi neerin g  13, 524-541. \ndoi:10.1109/TNSRE.2005 .857687 \nSzymanski, L. J., Kellis, S. , Liu, C. Y., Jon es, K. T., Ande rsen, R . A., Commins, D., e t a l. (2021). \nNeuro path ological effects of chronically i mplanted , intrac ortical micro elec trod es in a tet raplegic pa tien t. \nJour nal of Ne ural En gineeri ng  18, 0460b 9. doi:10.1088/1741-2552/ac127e \nThakore, V., M olnar, P. an d Hickman, J. J.  (2012). An Optimization-B ased Stu dy of Equivalent Circuit \nModels for Rep resen ting Reco rdings at t he Neu ron–El ectro de In terfac e. IEEE Tra nsact ions o n Biom edic al \nEngineeri ng  59, 2338-2347. doi:10 .1109/TBME.2012.2203820  \nVelliste, M ., Perel , S., Spal ding, M. C., W h itford, A . S. and Schwar tz, A . B. (2008). Cortical con trol of a \nprosthe tic arm for self-feeding. Nat ure  4 53, 1098-1101. doi:10.1038/natur e0699 6 \nWodlinger , B., Downey, J . E., Tyler-Kabar a, E. C., Schwartz , A. B ., Boning er, M . L. a nd Collinger, J. L. \n(2014). Ten-dimensional anthro pomorph ic arm control in a human br ain−machin e  interface : difficulties, \nsolutions, and limi tati ons. J ourn al of Ne u ral Engine ering  12, 016011 . doi:10.1088/ 1741-\n2560/12/1/01 6011  \nWoepp el, K., Yang, Q. a nd Cui, X. T. (2017). Recent advanc es in neural elect rode – tissue int erfaces. \nCurrent Op inio n in Bio medi cal En gineeri n g  4, 21-31. doi: ht tps://doi.o rg/10.1016/j.cobme.2017.09.003\n \nWoolley, A . J. , Desai, H. A . and O tt o, K. J . (2013). Chronic intracortical micro elect r ode arr ays induce non-\nuniform, depth-r ela ted tissu e respo nses.  Journ al of Neur al Engi neeri ng  10, 02600 7. doi:10.1088/1741-\n2560/10/2/02 6007  \nXie, X., Riet h, L., Willi ams, L., Negi, S ., Bh andari, R ., Caldwell, R ., et al. (2014). Long-term reliabili ty of \nAl2O3and Parylen e C bilayer encapsul ate d Utah el ectro de ar ray based ne ural in te rfaces for chronic \nimplanta tion. Jour nal of Ne ural En gineeri ng  11, 026016. doi:10 .1088/1741-2560/1 1/2/026016  \n \n \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\nTable 1. Number of electrically connected electrodes that were classified as \nundamaged/unencapsulated or damaged/encapsulated based on tip degradation, shank \ndegradation, and tissue encapsulation. \n Tip Degradation Shank Degradation Encapsulation \n Low (%) High (%) Low (%) High (%) Low (%) High (%) \nP1 122 (72.2) 47 (27.8) 151 (84.8) 25 (15.2) 51 (27.6) 134 (72.4) \nP2 Pt 146 (90.1) 16 (9.9) 172 (98.3) 3 (1.7) 90 (50.6) 88 (49.4) \nP2 IrOx \nMedial 26 (92.9) 2 (7.1) 32 (100) 0 (0) 6 (18.8) 26 (81.2) \nP2 IrOx \nLateral 7 (25) 21 (75) 9 (32.1) 19 (67.9) 30 (100) 0 (0) \nTotal 301 (77.7) 86 (22.3) 364 (88.8) 46 (11.2) 177 (41.6) 248 (58.4) \nExcluded 45 22 7 \n \nTable 2. Differences observed between patients with different length of implant (980 days for P1 \nand 182 days for P2) on material degradation and encapsulation for electrically connected \nplatinum recording electrode arrays \n P1 (%) P2 Pt (%) Fisher exact p-value \nDegraded Tips 27.8 9.9 <0.001 \nDegraded Shank 15.2 1.7 <0.001 \nEncapsulated 72.4 49.4 <0.001 \n \nTable 3. Effect of stimulation on material degradation and encapsulation for IrOx arrays. Non-\nstimulated tips were not electrically connected.  \n Stimulated (%)  Non-stimulated (%) Χ 2 Statistic p-value \nDamaged Tips 41.1 5.6 14.7 <0.001 \nDamaged Shank 33.3 0.0 19.3 <0.001 \nHigh Encapsulation  43.3 39.7 0.1 >0.05 \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\n \n \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\n \n  \n \nE\nG\nF\nH\nA B\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\n  \n  \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\n \n \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\n  \n \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\n \n \n \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\nSupplemental \n  \n \n \n \nLow Degradation/Unencapsulated \nHigh Degradation/Encapsulated \nExcluded from analysis \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\n  \n \n \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\n \n  \n \nA\n B\n C\nD\n E\n F\nG\n H\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\n \n  \n \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\n \n \n \n  \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\nFigure 1. Six electrode arrays were implanted in two participants: two recording arrays in \nP1 motor cortex, two recording arrays in P2 somatosensory cortex, and two stimulating \narrays in P2 medial parietal cortex. Intraoperative images of implanted arrays in P1 (A) and \nP2 (B). \nFigure 2 : Impedances and peak-to-peak voltages decreased over time.  Data points \nrepresent the median across electrodes for a given test date. The shaded regions show the \ninterquartile ranges smoothed with a nine-point moving average filter with a triangular kernel. \nMedian impedances recorded on (A) P1 electrodes and (B) P2 electrodes across the length of \nimplant. Impedance measurements on P1 were not conducted with the same temporal \nresolution as P2. Different colors represent platinum or IrOx for P2 as indicated in the legend. \nVpp recorded on (C) P1 electrodes and (D) P2 electrodes across the length of implant. For P1, \nthere was a discontinuity in the Vpp at day post-implant 550 due to a change in the RMS \nthreshold from -5.25 to -4.5. Overlayed impedances and Vpp for P1 and P2 are shown in (E) \nand (F), respectively. \nFigure 3: Characterization of the encapsulation of the electrodes. Arrays were imaged with \nan optical microscope in air. Both arrays are from P1. The encapsulation of array (A) was further \nexamined with TPM. (B) The location of 2P imaging along the Z axis and select electrode \nshanks. The array was stained for cell nuclei and zoomed-in images were taken of the green (C) \nand red (D) regions. In both regions there is prominent second harmonic signal, indicating the \npresence of collagen. The array in (E) was chosen to display the lack of homogeneity of the \nencapsulating tissues. (F) Location along the z-axis (blue box) and 2 selected areas further \nimaged. 3D rotation images were generated displaying the tissue encapsulation and nuclei \nstaining from the regions highlighted in green (G) or red (H). The outer image (G) displays high \nsecond harmonic signal while the inner image (H) has elevated cell counts, demonstrating the \nheterogeneity of the encapsulation. \nFigure 4: Brain vascularization can be visualized on one of the explanted arrays from P2. \nThe pre-implant location is indicated with a yellow box (A). (B) Optical image of the array \nshowing tissue coverage. (C) TPM of the shanks of the electrode, with green denoting second \nharmonic signal from collagen and red denoting the autofluorescence of the device. Each \nelectrode was imaged and separated by row (side view). Most of the electrode tips are covered \nby collagenous tissue. (D) TPM image of the array looking from the tips downward, with a \nportion of vasculature marked in blue. (E) zoomed in image from (A) where electrode shanks \nare superimposed on the underlying vasculature. (F) The vasculature visualized in (D) is \nsuperimposed on (E), showing similar trajectory, demonstrating that the vasculature structure \nidentified in the tissue on the explanted array is likely of pia origin.  \nFigure 5 . Tip and shank damage occurred on some implanted electrodes and \nencapsulation occurred on four implanted arrays. Representative high magnification images \nof undamaged/unencapsulated and damage/encapsulated electrodes. Tip images were taken \nfrom P1 array 1, with the degraded tip showing demetallation and biologic fouling (scale bar is \n10µm). Shank images were taken from P2 lateral stimulating array (scale bare is 100µm). The \ndegraded shank shows multiple surface and subsurface irregularities including pitting and \ndelamination from the tip. Encapsulation images were from P2 medial stimulating array (scale \nbare is 100µm). \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\nFigure 6. Stimulation-induced material damage on one of the two arrays. (A) SEM image of \nfour shanks of the lateral array, tip damages are found on the stimulated electrodes marked with \nwhite arrows. (B) EDS of the stimulating electrodes for the lateral array showing reduced \npresence of iridium (magenta) on most of the stimulated sites (white arrows). (C) SEM image of \nthe medial stimulating array tips. No differences were observed between the non-stimulated and \nstimulated tips on this array. Scale bars are 100µm. (D,E) Arrays showing the measured \nmaterial properties on the stimulation arrays including tip categories (D) and shank categories \n(E). Green spaces show electrodes categorized as undamaged/unencapsulated, blue spaces \nshow electrodes categorized as damaged, and black spaces show electrodes that were \nexcluded from analysis. (F,G). Medial (top) and lateral (bottom) stimulating arrays showing (F) \nminimum interphase voltage (G) and total charge injected below -0.6V. The color bar for the \nminimum interphase voltages is log-transformed to emphasize differences between electrodes. \nGrey spaces indicate unwired electrodes. White spaces indicate wired electrodes that were \nnever stimulated. (H) measured peak-to-peak voltages on stimulated electrodes after removing \nelectrodes which were encapsulated with fibrous tissues. There were no significant differences \nobserved in the measured unit amplitudes (Mann-Whitney non-parametric test) \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint \n\nSupplemental Figure 1: Categories assigned to each electrode site. Each site was \nassigned a category with respect to their material integrity or degree of encapsulation. Black \nsites were not able to be categorized and were excluded from the analysis.  \nSupplemental Figure 2 Optical images P2 arrays: A) Medial stimulating array from P2. B) \nMedial recording array from P2. C) Lateral recording array from P2. D) Lateral stimulating array \nfrom P2.  \nSupplemental Figure 3. Scanning Electron Microscopy reveals damage and \nencapsulation on a fraction of implanted electrodes. A,B) Recording arrays implanted into \nP1 after enzymatic treatment. C) Medial stimulating array implanted into P2. D,E) medial and \nlateral recording arrays implanted into P2. F) lateral stimulating array implanted into P2. G,H) \nHigher magnification images of recording array in (E) and stimulating arrays in (F), respectively. \nWhite arrows indicate electrodes which were used for stimulation. Red arrows indicate \nrepresentative electrodes which were excluded from analysis. \nSupplemental figure 4: Comparisons of impedance and Vpp for P1, P2 platinum \nelectrodes, and P2 IrOx electrodes on the last day of recording. No trends between \nimpedance and Vpp were observed for either of the platinum tipped recording arrays.  \nSupplemental Figure 5: Impedances and Vpp for individual electrodes on the last day of \nrecording. Lower impedances were observed on the arrays implanted in P1 due to the length of \nimplantation. Impedances were not measured on every electrode in P2 due to hardware \nlimitations, namely that the impedance cable could only measure impedances from a total of 96 \nchannels across two arrays. Colors are linearly scaled. \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 31, 2021. ; https://doi.org/10.1101/2021.08.28.21262765doi: medRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}