In the past two decades, the advancement of implanted electrode arrays has drastically shifted the field of neuroscience from theories based on individually recorded neurons to those based on the systems-level activity of neuron populations (Cunningham and Yu, 2014; Vyas et al., 2020; Ebitz and Hayden, 2021). Specifically, the 96-channel Utah array (Blackrock Neurotech, Salt Lake City, UT) has been used to make pivotal discoveries in motor systems neuroscience and develop clinically impactful brain-computer interfaces (Maynard et al., 1997; Serruya et al., 2002; Gilja et al., 2012; Churchland et al., 2012; Gallego et al., 2020; Vyas et al., 2020). These electrode arrays are widely used in neuroprosthetics research for patients with motor disorders and have been used for a variety of purposes, such as movement and speech decoding (Gilja et al., 2015; Willett et al., 2023). Much of this work in humans relies on the translation of results discovered in non-human primates (NHPs), a crucial animal model in neuroscience research (Roelfsema and Treue, 2014; Mitchell et al., 2018). One useful application of these arrays in animal studies is microstimulation and electrolytic lesion experiments (Tehovnik et al., 2006). Electrolytic lesioning results in small, controllable, and precise amounts of neuron loss. This makes it a helpful technique to study damage to targeted areas of the brain, serving as a causal tool for basic scientific inquiry and as a model for clinical injury-based neuron loss events at a small scale.

Electrolytic lesioning has been experimented with since the 19th century (Kline, 2007), and methods have been researched in many species, including rodents, cats, dogs, monkeys, and humans (Horsley and Clarke, 1908; Sweet, 1953; Chen et al., 2009). In electrolytic lesion experiments, researchers pass current through implanted electrodes in order to mark the position of electrodes or create lesions in brain tissue (Chen et al., 2009; Townsend et al., 2002). These electrolytic lesions can be created using one electrode (unipolar) or by using two electrodes (bipolar) (Sweet, 1953; Chehrazi and Collins, 1981). In particular, recent work has established a novel lesioning technique to create consistent, controllable electrolytic lesions in NHPs using already-implanted Utah arrays, without damaging the ability to record from the array (Bray et al., 2024). However, while recording capabilities seem to remain stable, the physical effect of electrolytic lesioning on the recording array needs further characterization.

Previous studies have shown temporally decreasing performance of collected signals from long-term implanted arrays, thought to be a consequence of the immune response, mechanical movement of the array relative to surrounding tissue, and physical degradation of the array (Suner et al., 2005; Simeral et al., 2011; Downey et al., 2018). This is also reflected in studies that specifically examine the damages that occur to electrode arrays while they are implanted in NHPs (Chestek et al., 2011; Chen et al., 2023). In particular, the use of scanning electrode microscopy (SEM) has been used to visually examine and analyze the effect of neural implantation on arrays with extremely high definition (Bjånes et al., 2024; Chen et al., 2023; Patel et al., 2023; Barrese et al., 2016; Woeppel et al., 2021). These studies, which span both NHPs and humans, provide direct evidence that the observable physical damage incurred by implanted arrays has direct impact on the array’s performance.

The Utah array is a silicon-based array manufactured with several layers of different materials (Yi et al., 2022). Silicon electrodes are first etched out of a glass and silicon base. Each individual electrode is then metallized at the tip. Finally, a layer of parylene C coats all but these metallized tips to help isolate and insulate each electrode (Campbell et al., 1991; Bhandari et al., 2010). Both in vitro and in vivo testing of electrode arrays revealed environmental damage to these materials, such as cracking, textural defects, and degradation in response to the brain’s temperature and salinity (Caldwell et al., 2020; Barrese et al., 2013). The immune response of the brain also damages the electrodes due to effects like glial scarring (gliosis) and inflammation (Polikov et al., 2005; Barrese et al., 2013; Salatino et al., 2017). This damage may be exacerbated by the surgical techniques used during implantation, which include pushing the electrode array into cortex and tethering the implant to the skull (Polikov et al., 2005; Biran et al., 2007; Potter et al., 2012). These results are confirmed in the previously mentioned SEM analyses (Barrese et al., 2016; Woeppel et al., 2021; Chen et al., 2023; Patel et al., 2023; Bjånes et al., 2024).

Additionally, the Utah array is available with both platinum and iridium oxide-based coatings. Aggressive electrical stimulation is known to dissolve platinum-based electrodes (Shepherd et al., 2021; Shah et al., 2024). Other studies have shown iridium oxide to be more resistant to stimulation-related damage, but not completely insusceptible (Cogan et al., 2004; Negi et al., 2010; Chen et al., 2014; Bjånes et al., 2024). However, the physical electrode damage associated with electrolytic lesioning remains to be studied.

This work comprises images of 11 different multielectrode Utah arrays (ten 96-channel, one 64-channel). Of these imaged arrays, eight arrays chronically implanted in regions of the motor cortex for varying amounts of time were analyzed further (Monkey H = 2242 days, F = 1875, U = 2680, C = 594). Four out of eight of these arrays were used to perform electrolytic lesioning experiments (n = 4 monkeys; H = 10 lesions, F = 1, U = 4, C = 1). Additionally, the image set includes images of a broken 96-channel Utah array, which was shattered during explant but remained intact while implanted. This set of images represents the largest publicly available collection of high-quality, single-electrode SEM images of explanted multielectrode Utah arrays. A total of 938 individual electrodes are available to view, along with 11 additional images from the broken array. Specific details for each imaged array are available in Supplementary file 1.

All collected images, along with their associated scores, are available in interactive Video 1. Each of the 12 arrays and individual electrodes may be selected and viewed. Representative electrodes for each category and damage score, as well as electrodes used for lesioning, are also viewable for each array in an additional drop-down menu when available. Rough anatomical layouts of the array’s implant location are also included. Electrode numbering maps are available in Video 1—source data 2.

Overall, these results collectively demonstrate that there is no obvious, statistically significant difference between observed damage to electrodes used versus not used for electrolytic lesioning. This indicates that any material effects potentially caused by electrolytic lesioning are indistinguishable from the typical damage seen in long-term implanted arrays. Because this method of lesioning involves passing weak direct current into the brain over relatively short time periods, its physical effects on the array electrodes are similar to alternating current stimulation. A recent study found that SEM-visible damage caused by stimulation on iridium oxide electrodes is highly variable between arrays, where one imaged array demonstrated noticeable stimulation-related damage and the other array had no such damage (Woeppel et al., 2021). Furthermore, the study found that array recording quality was not compromised by stimulation. Similarly, other work did not show significant differences in SEM-visible degradation between platinum electrodes used for stimulation and iridium oxide electrodes used for stimulation (Bjånes et al., 2024; Chen et al., 2023). All of these previous studies used different stimulation protocols: a pulse train delivered at 20–300 Hz at amplitudes 1–100 μA (Woeppel et al., 2021), a pulse train delivered at 300 Hz at amplitudes 1–210 μA (Chen et al., 2023), and an undisclosed low-current stimulation (Bjånes et al., 2024). Additionally, dissolution of electrodes due to current is most salient at aggressive levels of stimulation; one study reviewing electrode dissolution delivered variable amplitudes of current at 50 Hz for 7 hr Negi et al., 2010; another study utilized currents at least one order of magnitude greater than the above protocols (Shepherd et al., 2021). The electrolytic lesioning in this work primarily used direct currents delivered at around 150 μA for 45 s, though parameters ranged between 50–450 μA and 12–600 s. Thus, despite the differences between specific lesioning and stimulation protocols, the findings presented here are consistent with previous studies that found no unusual damage to the electrodes used for stimulation. This aligns with previous findings that electrolytic lesioning does not significantly impact the recording ability of electrodes (Bray et al., 2024).

The patterns of damage found in this work also largely reflect previous SEM analyses of explanted arrays. A prior study also found that the electrodes closest to the perimeter of the array suffered the most damage (Bjånes et al., 2024). This pattern is also visually observable in other published results (Patel et al., 2023; Woeppel et al., 2021). This is likely because the edges of the arrays serve as a physical shield for the innermost electrodes. Edge electrodes are subject to greater access to brain tissue, as well as to mechanical damage incurred during surgery and handling.

The frequency of damage types identified in this work differs from previous studies, but is not unusual. In this work, the most common damage type within arrays used for lesioning is abnormal debris (n = 276, 71.9%), followed by metal coating cracks (n = 214, 55.7%), tip breakage (n = 102, 26.6%), parylene C cracks (n = 95, 24.7%), and parylene C delamination (n = 48, 12.5%). On these lesioning arrays, there were n = 53, or 13.8%, shank fractures. Similarly, across all eight NHP arrays imaged, the most common damage type is abnormal debris (n = 518, 67.4%), followed by metal coating cracks (n = 480, 62.5%), tip breakage (n = 233, 30.3%), parylene C cracks (n = 224, 29.2%), and parylene C delamination (n = 60, 7.8%). Across all eight NHP arrays, there were n = 117, or 15.2%, shank fractures. However, a prior study found that coating cracks were the most common type of observed damage, followed closely by parylene C cracks, then tip breakage, abnormal debris, and parylene C delamination (Patel et al., 2023). While the frequency of damage types is not exactly aligned, previous work included a cleaning step during array preparation for SEM imaging; this work does not explicitly attempt to remove excess brain tissue from imaged arrays, which likely led to the relative increase we observed in abnormal debris. Additionally, variation in individual researchers’ scoring standards may also contribute to differences. As mentioned above, the overall spatial pattern of damage between these two studies is consistent. Finally, this study also found a moderate positive correlation between coating cracks and tip breakage. This reflects the fact that damage to the silicon core of an electrode often occurs in tandem with cracking and flaking of the surrounding metal coating; the silicon cannot break without also breaking its coating. This correlation is visible in the collected images presented in Video 1.

This work analyzed high-definition SEM images of previously implanted electrode arrays used for electrolytic lesioning. Across all eight arrays previously implanted in NHPs, damage disproportionately occurred to the outer edges of arrays. Additionally, no statistically significant difference was found between the damage experienced by normal and lesioning electrodes within the same array. Furthermore, statistical testing between arrays used and not used for lesioning experiments did not indicate consistent, significant differences. Finally, this work also presents the largest publicly available set of SEM images of explanted arrays, consisting of 11 different multielectrode Utah arrays (ten 96-channel, one 64-channel) and one broken 96-channel array. These results from this dataset align with previous SEM studies of chronically implanted arrays used for both recording and stimulation, while providing further evidence for electrolytic lesioning as a safe and useful technique for experimental neuroscience.