Brain–computer interfaces re-shape functional neurosurgery

Invasive as well as non-invasive neurotechnologies conceptualized to interface the central and peripheral nervous system have been probed for the past decades, which refer to electroencephalography, electrocorticography and microelectrode arrays. The challenges of these mentioned approaches are characterized by the bandwidth of the spatiotemporal resolution, which in turn is essential for large-area neuron recordings (Abiri et al., 2019). In a most recent consensus statement, the brain-computer-interface (BCI) society defined BCIs as a system that measures brain activity and converts it in (nearly) real-time into functionally useful outputs to replace, restore, enhance, supplement, and/or improve the natural outputs of the brain, thereby changing the ongoing interactions between the brain and its external or internal environments. It may additionally modify brain activity using targeted delivery of stimuli to create functionally useful inputs to the brain (Soldado-Magraner et al., 2023). Nascent developments have enabled the conceptualization of small-scale microelectrode arrays, which more and more are probed to couple electrophysiological patterns of brain circuits with internal and/or external environmental properties (e.g., prosthetics, computer) seeking to restore an impaired state of motor performance (Hochberg et al., 2012; Collinger et al., 2013; Bouton et al., 2016; Hughes et al., 2021; Patrick-Krueger et al., 2024). At this point, it is noteworthy that despite the BCI array itself, additional equipment and an interdisciplinary approach (including but not limited to neurophysiology, bioengineering, and computational science) are mandatory to ensure proper signal acquisition, signal processing, feature extraction, and pattern recognition which in turn may control external functionally used prosthetics, wheelchairs, monitors, and speech speller (Patrick-Krueger et al., 2024).

This perspective aimed to discuss the current state of invasive surgically implanted BCI microelectrode arrays for motor recovery from a functional neurosurgical perspective as the numbers of surgically implanted BCI technologies more and more have been integrated into the neurosurgical theatre. An additional purpose was to showcase and discuss a combined framework on how adjunct peripheral non-invasive tools like high-density surface electromyography (HD-sEMG) may advance the use of BCI technologies to restore upper limb/hand motor tasks.

Spinal cord injured individuals may achieve a markedly improved and sustained quality of life level through different rehabilitative therapies; however, survey-based studies indicate that spinal cord injured patients suffer from motor impairment affecting subdomains of quality of life associated with the loss of their ability to perform manual tasks crucial for daily execution. Given these data, recovering hand function is one of their top priorities, being comparable to bladder and bowel functions. Even partially improving hand function means more independence for these individuals and reduced healthcare costs since less attendant care is required (Snoek et al., 2004).

Therapeutic approaches to recover hand movements include hand reconstructive surgery, neuromodulation, the use of neuroprosthetics, assistive devices, and neurorestorative therapies. Surgical interventions include tendon or nerve transfer, and neurorestorative therapies for spinal cord injured recovery remain in an experimental stage, and current treatment approaches are still far from being translated into the realm of clinical practical use (Oliveira et al., 2024). Central and peripheral invasive BCIs gained increased attention, nowadays representing a promising field for restoring motor impairment by linking the nervous system with the environment (e.g., assistive devices) and characterized by the ability to record cortical or peripheral circuits with the main goal to translate these signals into a functional motor/behavioral output (Abiri et al., 2019; Patrick-Krueger et al., 2024).

Clinical research lines indicate that intracortical implanted microarrays (BCI) may provide sustained movement control for tetraplegic subjects with a deteriorated movement output, demonstrating the potential for functional restoration of the upper limbs/hand (Abiri et al., 2019; Patrick-Krueger et al., 2024). Invasive cortical interfaces have the advantage to present signals with high resolution and signal-to-noise ratio with the capability to detect single neuron spikes sorted from mixed signals of a large neuron population. However, the signal quality tends to decline due to host-graft interactions (e.g., gliosis) over time. Several reports of in-human applications for motor recovery of the upper limb have targeted the brain (e.g., motor cortex) as well as the peripheral nerve system, yielding encouraging results covering an extended observation period (Bockbrader et al., 2019; Vu et al., 2020; Page et al., 2021).

In comparison, non-invasive methods usually present with a lower signal resolution and signal-to-noise ratio, although being characterized by task-specific and pattern-relevant motor behaviors (e.g., electroencephalography). However, most recent human studies indicate that patients with tetraplegia (C₄–C₆) can still have residual motor neuron activity, which can be reliably detected using HD-sEMG and associated processing algorithms tailored for this type of motoneuron signaling. However, HD-sEMG may be limited by spatial-temporal resolution, non-stationarity of signals, generalizability, and task constraints compared to an invasive central BCI implant during combined precision and postural tasks, such as concurrent grasping of objects, holding, and modulating the frequency of movement (Oliveira et al., 2024). Consecutively, the entire neural signal chain may not be mapped sufficiently with purely peripheral non-invasive HD-sEMG measurements, therefore building the foundation to study the spared corticospinal connections between the motor cortex and spinal motor neurons at the direct central levels (Figure 1).

Figure 1.

Schematic illustration for the implementation of central invasive (BCI array) and peripheral non-invasive neural interfaces (HD-EMG) in human subjects with motor impairment and beyond.

The recordings of thousands of cortical and spinal motoneurons in a fully automatic way will tackle the most difficult questions such as how the human cortex and spinal cord coordinate highly skilled motor behaviors. By recording, the relationship between the activity of cortical neurons (in the brain) and motor neurons in the spinal cord may lead to the development of more effective hybrid neural interfaces to improve the precision and accuracy of prosthetic devices and other assistive technologies that rely on neural signals for control. BCI: Brain–computer interface; HD-EMG: high-density electromyography.

In this way, Oliveira et al. (2024) proposed a combined approach applying peripheral non-invasive and central invasive neural interfaces, which, in the first step, recorded spared signals from the spinal cord with minimally invasive as well as non-invasive sensors, either implanted in the muscle or recording motor unit action potential from the skin using a non-invasive HD-sEMG interface. As current peripheral non-invasive interfaces lack selectivity for studying individual neurons, using an invasive intracortical microelectrode array is deemed essential as proposed by Oliveira et al. (2024) to appropriately acquire and analyze the signaling pathways relevant for the execution of specific motor tasks in real-time. Moreover, coupling cortical recordings with spinal motor neuron activity would permit the establishment of a hybrid classifier of motor intention, enabling a correlation between movement intention and muscle activity (Figure 1).

This proposed concept by Oliveira et al. (2024) aimed to increase the reliability of the acquired signals while maintaining a high signal resolution, improve the performance (less computational power required due to redundancy and increased information transfer), and overcome non-stationarity (variation in brain signals) and ultimately improving the spatiotemporal decoding of the signals (Figure 1). A previous study has shown that motoneurons are spared in many neurological conditions, such as clinically complete and incomplete spinal cord injury, stroke, and motoneuron disease (Oliveira et al., 2024). In seeking the electrophysiological clues that arise from the brain and permit its executive communication with peripheral motor units, Oliveira et al. have recently reported the presence of spared spinal motoneuron activity below the level of lesion modulated by central motor commands in a relatively large group of C₄–C₆ tetraplegic individuals during attempted hand and digit movements, as these results provide evidence of voluntarily controlled spinal motor neurons in spinal cord injury. The same working group further demonstrated in a previous human study that even a small subset of spinal motoneurons, accessed in an entirely non-invasive way via HD-sEMG, allows the prediction of relevant neural dynamics associated with movement in young, healthy individuals and new-borns (Oliveira et al., 2024). The concurrent recordings of spared spinal motoneurons and cortical data in real time will provide unprecedented information on how the cortex controls muscle forces in humans. These additional inputs will be essential for automatic scaling and calibration of cortical input into movement kinetics since spared motoneurons are scaled proportionally to muscle forces and can be obtained in real time with the proposed method, which in turn is mandatory for a reliable information transfer from the central and peripheral nervous system to assistive devices.

Neuromodulation techniques rely on electrical, light, magnetic, ultrasound, and other stimulation methods used to interfere with pathological circuits at the brain level, spinal cord, and peripheral nerve system for the treatment of a broad range of neurological disorders. It is worthy to briefly mention that with respect to clinically established and intracranially implanted brain stimulation devices such as deep brain stimulation, BCIs (temporarily and/or permanently) can inform deep brain stimulation to adapt stimulation waveforms correlative to the electrophysiological signatures of pathological circuits (Zhu et al., 2024). In this way, artificial intelligence can be used to train a neuronal network to predict movements and/or to detect pathological electrophysiological patterns relevant to the applied stimulation waveforms.

Despite the recent burgeoning interest and concern about the use of invasive BCI neuro-applications aiming to unlock unresolved neurological therapeutic gaps, a neuroethical framework posing ethical, legal, social, and cultural implications is urgently needed and should ideally prelude the regular implementation of such neurotechnologies. The purpose of the ethical, legal, social, and cultural framework remains undefined. Such a framework aims to identify and characterize ethical, legal, social, and cultural issues relevant for research, development, clinical application, evaluation, and usage of BCI technologies, ultimately proposing mitigation measures and policy options to maximize its benefits while minimizing its risks. Nevertheless, potential drawbacks that prevent BCI technologies from being used broadly are high costs, open questions related to reimbursement from healthcare providers, the use/storage of sensitive brain data, and the necessity of permanent brain implants.

In view of the fact that this exciting field is in the process of maturation, it is of high importance to recognize that the use of BCIs is and will not be restricted only to medical applications but will find its way into potential additional fields far beyond medicine, such as wellness, work/employment, education, entertainment, marketing, sports, and military (Soldado-Magraner et al., 2024).