Exploring Cognition with Brain-Machine Interfaces

Neural signatures of temporal anticipation in human cortex represent event probability density

Temporal prediction is a fundamental function of neural systems. Recent results show that humans anticipate future events by calculating probability density functions, rather than hazard rates. However, direct neural evidence for this hypothesized mechanism is lacking. We recorded neural activity using magnetoencephalography as participants anticipated auditory and visual events distributed in time. We show that temporal anticipation, measured as reaction times, approximates the event probability density function, but not hazard rate. Temporal anticipation manifests as spatiotemporally patterned activity in three anatomically and functionally distinct parieto-temporal and sensorimotor cortical areas. Each of these areas revealed a marked neural signature of anticipation: Prior to sensory cues, activity in a specific frequency range of neural oscillations, spanning alpha and beta ranges, encodes the event probability density function. These neural signals predicted reaction times to imminent sensory cues. These results demonstrate that supra-modal representations of probability density across cortex underlie the anticipation of future events.

Quantifying physical degradation alongside recording and stimulation performance of 980 intracortical microelectrodes chronically implanted in three humans for 956-2130 days

The clinical success of brain computer interfaces (BCI) depends on overcoming both biological and material challenges to ensure a long-term stable connection for neural recording and stimulation. This study systematically quantified damage that microelectrodes sustained during chronical implantation in three people with tetraplegia for 956-2130 days. Using scanning electron microscopy (SEM), we imaged 980 microelectrodes from eleven Neuroport arrays tipped with platinum (Pt, n = 8) and sputtered iridium oxide film (SIROF, n = 3). Arrays were implanted/explanted from posterior parietal, motor and somatosensory cortices across three clinical sites (Caltech/UCLA, Caltech/USC, APL/Johns Hopkins). From the electron micrographs, we quantified and correlated physical damage with functional outcomes measured in vivo, prior to explant (recording quality, noise, impedance and stimulation ability). Despite greater physical degradation, SIROF electrodes were twice as likely to record neural activity than Pt (measured by SNR). For SIROF, 1 kHz impedance significantly correlated with all physical damage metrics, recording metrics, and stimulation performance, suggesting a reliable measurement of in vivo degradation. We observed a new degradation type, primarily on stimulated electrodes ("pockmarked" vs "cracked") electrodes; however, no significant degradation due to stimulation or amount of charge delivered. We hypothesize erosion of the silicon shank accelerates damage to the electrode / tissue interface, following damage to the tip metal. These findings link quantitative measurements to the microelectrodes' physical condition and their capacity to record/stimulate. These data could lead to improved manufacturing processes or novel electrode designs to improve long-term performance of BCIs, making them vitally important as multi-year clinical trials of BCIs are becoming more common. STATEMENT OF SIGNIFICANCE: Long-term performance stability of the electrode-tissue interface is essential for clinical viability of brain computer interface (BCI) devices; currently, materials degradation is a critical component for performance loss. Across three human participants, ten micro-electrode arrays (plus one control) were implanted for 956-2130 days. Using scanning electron microscopy (SEM), we analyzed degradation of 980 electrodes, comparing two types of commonly implanted electrode tip metals: Platinum (Pt) and Sputtered Iridium Oxide Film (SIROF). We correlated observed degradation with in vivo electrode performance: recording (signal-to-noise ratio, noise, impedance) and stimulation (evoked somatosensory percepts). We hypothesize penetration of the electrode tip by biotic processes leads to erosion of the supporting silicon core, which then accelerates further tip metal damage. These data could lead to improved manufacturing processes or novel electrode designs towards the goal of a stable BCI electrical interface, spanning a multi-decade participant lifetime.

Population encoding of observed and actual somatosensations in the human posterior parietal cortex

Cognition relies on transforming sensory inputs into a generalizable understanding of the world. Mirror neurons have been proposed to underlie this process, mapping visual representations of others' actions and sensations onto neurons that mediate our own, providing a conduit for understanding. However, this theory has limitations. Here, we hypothesize that mirror-like responses represent one facet of a broader framework in which our brains engage internal models for cognition. We recorded populations of single neurons in the human posterior parietal cortex (PPC) of a brain-machine interface clinical trial participant implanted with a microelectrode array while she either experienced actual touch, or observed diverse tactile stimuli applied to other individuals. Two body locations were tested, on each of the participant and other individuals. Some neurons exhibited mirror-like properties, consistent with earlier literature. However, they were fragile, breaking with increased task complexity. Population responses were better characterized by generalizable and compositional basic-level features encoded within neural subspaces. These features enable the population to respond to diverse actual and observed touch stimuli and are recruited similarly for similar forms of touch. Mirror-like neurons belong within these subspaces, contributing more globally to compositionality and generalizability. We speculate that at a population-level, human PPC manifests an internal model for touch, and that cognition unfolds in the high-level human cortex by versatility in its representational building blocks. In a broad sense, we speculate that the population features we demonstrate support a broad mechanism by which the high-level human cortex enables understanding.

Neural interfaces: Bridging the brain to the world beyond healthcare

Neural interfaces, emerging at the intersection of neurotechnology and urban planning, promise to transform how we interact with our surroundings and communicate. By recording and decoding neural signals, these interfaces facilitate direct connections between the brain and external devices, enabling seamless information exchange and shared experiences. Nevertheless, their development is challenged by complexities in materials science, electrochemistry, and algorithmic design. Electrophysiological crosstalk and the mismatch between electrode rigidity and tissue flexibility further complicate signal fidelity and biocompatibility. Recent closed-loop brain-computer interfaces, while promising for mood regulation and cognitive enhancement, are limited by decoding accuracy and the adaptability of user interfaces. This perspective outlines these challenges and discusses the progress in neural interfaces, contrasting non-invasive and invasive approaches, and explores the dynamics between stimulation and direct interfacing. Emphasis is placed on applications beyond healthcare, highlighting the need for implantable interfaces with high-resolution recording and stimulation capabilities.

Quantifying physical degradation alongside recording and stimulation performance of 980 intracortical microelectrodes chronically implanted in three humans for 956-2246 days

Motivation

The clinical success of brain-machine interfaces depends on overcoming both biological and material challenges to ensure a long-term stable connection for neural recording and stimulation. Therefore, there is a need to quantify any damage that microelectrodes sustain when they are chronically implanted in the human cortex.

Methods

Using scanning electron microscopy (SEM), we imaged 980 microelectrodes from Neuroport arrays chronically implanted in the cortex of three people with tetraplegia for 956-2246 days. We analyzed eleven multi-electrode arrays in total: eight arrays with platinum (Pt) electrode tips and three with sputtered iridium oxide tips (SIROF); one Pt array was left in sterile packaging, serving as a control. The arrays were implanted/explanted across three different clinical sites surgeries (Caltech/UCLA, Caltech/USC and APL/Johns Hopkins) in the anterior intraparietal area, Brodmann's area 5, motor cortex, and somatosensory cortex.Human experts rated the electron micrographs of electrodes with respect to five damage metrics: the loss of metal at the electrode tip, the amount of separation between the silicon shank and tip metal, tissue adherence or bio-material to the electrode, damage to the shank insulation and silicone shaft. These metrics were compared to functional outcomes (recording quality, noise, impedance and stimulation ability).

Results

Despite higher levels of physical degradation, SIROF electrodes were twice as likely to record neural activity than Pt electrodes (measured by SNR), at the time of explant. Additionally, 1 kHz impedance (measured in vivo prior to explant) significantly correlated with all physical damage metrics, recording, and stimulation performance for SIROF electrodes (but not Pt), suggesting a reliable measurement of in vivo degradation.We observed a new degradation type, primarily occurring on stimulated electrodes ("pockmarked" vs "cracked") electrodes; however, tip metalization damage was not significantly higher due to stimulation or amount of charge. Physical damage was centralized to specific regions of an array often with differences between outer and inner electrodes. This is consistent with degradation due to contact with the biologic milieu, influenced by variations in initial manufactured state. From our data, we hypothesize that erosion of the silicon shank often precedes damage to the tip metal, accelerating damage to the electrode / tissue interface.

Conclusions

These findings link quantitative measurements, such as impedance, to the physical condition of the microelectrodes and their capacity to record and stimulate. These data could lead to improved manufacturing or novel electrode designs to improve long-term performance of BMIs making them are vitally important as multi-year clinical trials of BMIs are becoming more common.

Neural subspaces of imagined movements in parietal cortex remain stable over several years in humans

Objective.A crucial goal in brain-machine interfacing is the long-term stability of neural decoding performance, ideally without regular retraining. Long-term stability has only been previously demonstrated in non-human primate experiments and only in primary sensorimotor cortices. Here we extend previous methods to determine long-term stability in humans by identifying and aligning low-dimensional structures in neural data.Approach.Over a period of 1106 and 871 d respectively, two participants completed an imagined center-out reaching task. The longitudinal accuracy between all day pairs was assessed by latent subspace alignment using principal components analysis and canonical correlations analysis of multi-unit intracortical recordings in different brain regions (Brodmann Area 5, Anterior Intraparietal Area and the junction of the postcentral and intraparietal sulcus).Main results.We show the long-term stable representation of neural activity in subspaces of intracortical recordings from higher-order association areas in humans.Significance.These results can be practically applied to significantly expand the longevity and generalizability of brain-computer interfaces.Clinical TrialsNCT01849822, NCT01958086, NCT01964261.

Similar neural states, but dissimilar decoding patterns for motor control in parietal cortex

Discrete neural states are associated with reaching movements across the fronto-parietal network. Here, the Hidden Markov Model (HMM) applied to spiking activity of the somato-motor parietal area PE revealed a sequence of states similar to those of the contiguous visuomotor areas PEc and V6A. Using a coupled clustering and decoding approach, we proved that these neural states carried spatiotemporal information regarding behaviour in all three posterior parietal areas. However, comparing decoding accuracy, PE was less informative than V6A and PEc. In addition, V6A outperformed PEc in target inference, indicating functional differences among the parietal areas. To check the consistency of these differences, we used both a supervised and an unsupervised variant of the HMM, and compared its performance with two more common classifiers, Support Vector Machine and Long-Short Term Memory. The differences in decoding between areas were invariant to the algorithm used, still showing the dissimilarities found with HMM, thus indicating that these dissimilarities are intrinsic in the information encoded by parietal neurons. These results highlight that, when decoding from the parietal cortex, for example, in brain machine interface implementations, attention should be paid in selecting the most suitable source of neural signals, given the great heterogeneity of this cortical sector.

Machine learning decoding of single neurons in the thalamus for speech brain-machine interfaces

Objective. Our goal is to decode firing patterns of single neurons in the left ventralis intermediate nucleus (Vim) of the thalamus, related to speech production, perception, and imagery. For realistic speech brain-machine interfaces (BMIs), we aim to characterize the amount of thalamic neurons necessary for high accuracy decoding.Approach. We intraoperatively recorded single neuron activity in the left Vim of eight neurosurgical patients undergoing implantation of deep brain stimulator or RF lesioning during production, perception and imagery of the five monophthongal vowel sounds. We utilized the Spade decoder, a machine learning algorithm that dynamically learns specific features of firing patterns and is based on sparse decomposition of the high dimensional feature space.Main results. Spade outperformed all algorithms compared with, for all three aspects of speech: production, perception and imagery, and obtained accuracies of 100%, 96%, and 92%, respectively (chance level: 20%) based on pooling together neurons across all patients. The accuracy was logarithmic in the amount of neurons for all three aspects of speech. Regardless of the amount of units employed, production gained highest accuracies, whereas perception and imagery equated with each other.Significance. Our research renders single neuron activity in the left Vim a promising source of inputs to BMIs for restoration of speech faculties for locked-in patients or patients with anarthria or dysarthria to allow them to communicate again. Our characterization of how many neurons are necessary to achieve a certain decoding accuracy is of utmost importance for planning BMI implantation.

Orbitofrontal high-gamma reflects spike-dissociable value and decision mechanisms

The orbitofrontal cortex (OFC) plays a crucial role in value-based decision-making. While previous research has focused on spiking activity in OFC neurons, the role of OFC local field potentials (LFPs) in decision-making remains unclear. LFPs are important because they can reflect synaptic and subthreshold activity not directly coupled to spiking, and because they are potential targets for less invasive forms of brain-machine interface (BMI). We recorded LFPs and spiking activity using multi-channel vertical probes while monkeys performed a two-option value-based decision-making task. We compared the value- and decision-coding properties of high-gamma range LFPs (HG, 50-150 Hz) to the coding properties of spiking multi-unit activity (MUA) recorded concurrently on the same electrodes. Results show that HG and MUA both represent the values of decision targets, and that their representations have similar temporal profiles in a trial. However, we also identified value-coding properties of HG that were dissociable from the concurrently-measured MUA. On average across channels, HG amplitude increased monotonically with value, whereas the average value encoding in MUA was net neutral. HG also encoded a signal consistent with a comparison between the values of the two targets, a signal which was much weaker in MUA. In individual channels, HG was better able to predict choice outcomes than MUA; however, when simultaneously recorded channels were combined in population-based decoder, MUA provided more accurate predictions than HG. Interestingly, HG value representations were accentuated in channels in or near shallow cortical layers, suggesting a dissociation between neuronal sources of HG and MUA. In summary, we find that HG signals are dissociable from MUA with respect to cognitive variables encoded in prefrontal cortex, evident in the monotonic encoding of value, stronger encoding of value comparisons, and more accurate predictions about behavior. High-frequency LFPs may therefore be a viable - or even preferable - target for BMIs to assist cognitive function, opening the possibility for less invasive access to mental contents that would otherwise be observable only with spike-based measures.

Neural subspaces of imagined movements in parietal cortex remain stable over several years in humans

A crucial goal in brain-machine interfacing is long-term stability of neural decoding performance, ideally without regular retraining. Here we demonstrate stable neural decoding over several years in two human participants, achieved by latent subspace alignment of multi-unit intracortical recordings in posterior parietal cortex. These results can be practically applied to significantly expand the longevity and generalizability of future movement decoding devices.

Implicit mechanisms of intention

High-level cortical regions encode motor decisions before or even absent awareness, suggesting that neural processes predetermine behavior before conscious choice. Such early neural encoding challenges popular conceptions of human agency. It also raises fundamental questions for brain-machine interfaces (BMIs) that traditionally assume that neural activity reflects the user's conscious intentions. Here, we study the timing of human posterior parietal cortex single-neuron activity recorded from implanted microelectrode arrays relative to the explicit urge to initiate movement. Participants were free to choose when to move, whether to move, and what to move, and they retrospectively reported the time they felt the urge to move. We replicate prior studies by showing that posterior parietal cortex (PPC) neural activity sharply rises hundreds of milliseconds before the reported urge. However, we find that this "preconscious" activity is part of a dynamic neural population response that initiates much earlier, when the participant first chooses to perform the task. Together with details of neural timing, our results suggest that PPC encodes an internal model of the motor planning network that transforms high-level task objectives into appropriate motor behavior. These new data challenge traditional interpretations of early neural activity and offer a more holistic perspective on the interplay between choice, behavior, and their neural underpinnings. Our results have important implications for translating BMIs into more complex real-world environments. We find that early neural dynamics are sufficient to drive BMI movements before the participant intends to initiate movement. Appropriate algorithms ensure that BMI movements align with the subject's awareness of choice.