Brain‐computer interfaces and plasticity of the human nervous system

Abstract

A brain-computer interface (BCI) is a system that processes signals associated with brain activity, such as EEGs, in order to control external devices, such as robotic assistance, without the use of nerves or muscles (Vidal, 1973). Development of BCIs has been addressed from multiple perspectives as several disciplines are relevant for this technology. In this special issue of The Journal of Physiology, we have focused attention on the physiological knowledge needed to further advance BCI technology in relation to the changes induced by BCIs on the central nervous system (CNS). Understanding the physiological mechanisms of plasticity induced by BCI systems is the basis of effective design of clinically viable therapeutic approaches. BCIs have been classically developed as means to re-establish communication or to restore movements in severely paralysed individuals, with a main focus on decoding user intention. Since the early developments of these approaches, it has become evident that the use of a BCI corresponds to learning a new motor or cognitive task and is therefore associated with learning and plasticity of the CNS. The neurophysiological bases of learning are, for example, the foundation of co-adaptation of the user with the machine learning algorithms developed to decode brain activity. Most BCI systems require extensive training by the user which follows the mechanisms of learning by the CNS. The crucial role of learning in the use of BCIs is exemplified by the development of BCI systems for patients who may progressively transition from a partial paralysis to a complete locked-in state (CLIS), such as patients with amyotrophic lateral sclerosis (ALS). ALS patients may be trained to the use of BCIs before entering the CLIS, with the expectation that communication will be maintained by the use of the same BCI systems when entering into CLIS. The passage from incomplete locked-in-state (LIS) to CLIS is critical for this clinical application of BCIs since it is fundamental to maintain information transfer when no other means of communication is possible. For this reason, understanding the neurophysiological factors determining BCI performance over time and its relations to learning is a central aspect of this clinical application. In this special issue Chaudhary et al. (2021) review the neuropsychological and neurophysiological aspects of the use of BCI in paralysis, specifically in ALS, and provide a perspective on the need to fill gaps in neurophysiological knowledge to advance these BCI systems. Learning a task mediated by a BCI is also associated with agency, i.e. the attribution of actions to ourselves. This is an important aspect in the use of BCIs for neurological rehabilitation, for example, in immersive virtual reality systems controlled by BCIs. The sense of agency induced by BCIs depends on the decoded brain signals and on the machine learning algorithm used for the decoding. These aspects of BCI research are analysed in the paper by Nierula et al. (2021), which compares the sense of agency and responsibility over actions arising from different types of BCIs. The results indicate that in applications in neurorehabilitation, a BCI may need to be designed taking into account the embodiment induced on the controlled external devices. In addition to developments for restoring communication or partial movements in fully paralysed individuals, a variety of BCI-based approaches have been proposed as neuromodulatory interventions to enhance recovery of motor and cognitive functions following neurological injuries, such as stroke. Although the clinical benefits of these approaches with respect to classic therapeutic interventions have been documented in a relatively small number of studies, therapeutic applications of BCIs are promising in a variety of conditions, for example in the initial phases of rehabilitation of stroke patients who cannot perform voluntary movements. Therapeutic applications of BCIs are based on