With the continuous development of artificial intelligence technology, “brain-computer interfaces” are gradually entering the field of medical rehabilitation. As a result, brain-computer interfaces (BCIs) have been included in many countries’ strategic plans for innovating this field, and subsequently, major funding and talent have been invested in this technology. In neurological rehabilitation for stroke patients, the use of BCIs opens up a new chapter in “top-down” rehabilitation. In our study, we first reviewed the latest BCI technologies, then presented recent research advances and landmark findings in BCI-based neurorehabilitation for stroke patients. Neurorehabilitation was focused on the areas of motor, sensory, speech, cognitive, and environmental interactions. Finally, we summarized the shortcomings of BCI use in the field of stroke neurorehabilitation and the prospects for BCI technology development for rehabilitation.
According to WHO clinical criteria, stroke is defined as “a rapidly developing sign of focal (or global) brain dysfunction lasting more than 24 hours (unless interrupted by death), with no apparent nonvascular cause.” Stroke is the world’s second leading cause of death and third leading cause of injury and can cause severe cognitive, emotional, and sensorimotor impairment in patients [
In traditional rehabilitation, the gold standard in care for poststroke recovery is a combination of specialized training and general aerobic exercise. Bimanual arm training (BAT) and constraint-induced movement therapy (CIMT) are two of the most established methods for treating stroke-related sports injuries [
The remodeling of neurological function after stroke may facilitate the development of new interventions for poststroke rehabilitation, and recent therapeutic options have shifted to facilitating neural circuit reorganization in order to restore motor function. These top-down approaches to rehabilitation are largely due to the mechanisms of brain plasticity [
With the advancement of science and technology, artificial intelligence technologies, such as brain-computer interface (BCI), virtual reality (VR), and augmented reality (AR), are rapidly developing and are gradually being applied in the field of medicine. Due to its direct action on the brain, BCI induces brain plasticity and promotes functional reorganization of the brain, proving to be a superior approach in poststroke rehabilitation, especially for improving motor function in stroke patients. The limited neurorehabilitation modalities are no longer adequate to meet increasing rehabilitation needs of patients with central injuries, and BCI has been shown to be effective in improving motor function and enhancing the lives of stroke patients. In this review, we first examined the latest BCI technologies, including how BCIs are acquired, how signals are processed, and how other artificial intelligence technologies are combined with BCIs, such as functional electrical stimulation (FES) technology, virtual reality, exoskeletons, orthotics, and intelligent wheelchairs. We then presented the specific applications, mechanisms of action, and efficacy of BCI in the treatment of poststroke neural remodeling, such as in BCI-based neurorehabilitation of stroke patients in motor, sensory, verbal, cognitive, and environmental interactions. Finally, we summarized our recent research findings and shortcomings, as well as an outlook on the development of BCI technology in the field of rehabilitation.
The word “brain-computer interface” was first formally identified as “a communication device that does not depend on the usual output pathways of the peripheral nerves and muscles of the brain” at the First International Conference on Brain-Computer Interface Technology in June 1999 [
Brain-computer interface for the acquisition, extraction, and conversion of signals from the brain for the ultimate application of controlling external devices: virtual reality, functional electrical stimulation, exoskeleton robots, and intelligent wheelchairs.
A BCI uses signals from the brain to gather information about the user’s intentions. To do this, the BCI relies on a recording phase to measure brain activity and will then convert that information into an electrical signal that can be easily processed. Depending on the BCI’s level of invasiveness, there are two types of recording methods: invasive and noninvasive. Invasive recording methods have more spatial and temporal precision, but they also come with the dangers that come with surgically implantable instruments. Extensive research into noninvasive recording techniques has quickly increased due to the technique’s noninvasive and safe nature. Due to the low quality of collected signals and susceptibility to interference, enhancing the signal quality of noninvasive brain-computer interfaces has become a focus of research. Depending on the form of signal acquisition, noninvasive BCIs are divided into electroencephalography (EEG), electrocorticography (ECoG), magnetoencephalography (MEG), intracortical electrical signal mapping (INR), near-infrared spectroscopy (NIRS), functional magnetic resonance imaging (fMRI), and many more. The classification of these acquired signals relies on two types of brain activity: (I) electrophysiological and (II) hemodynamic. These neural signal acquisition methods differ mainly in terms of activity detection, temporal and spatial resolution, safety, and mobility [
Comparison of BCI pivot signal acquisition methods and their advantages and disadvantages.
Nerve-signal acquisition methods | Event monitoring | Time resolution | Spatial resolution | Safety | Advantage | Disadvantage |
---|---|---|---|---|---|---|
Electroencephalogram (EEG) | Electrical signals | ~0.05 s | ~10 mm | Noninvasive | High temporal resolution, relatively low cost, high portability, low risk to users | Poor signal quality |
Magnetoencephalography | Magnetic signals | ~0.05 s | ~5 mm | Noninvasive | High temporal and spatial resolution, less training time, and more reliable communication | Technology is too large and expensive |
Electrocorticography | Electrical signals | ~0.003 s | ~1 mm | Invasive | High temporal and spatial resolution and low artefact vulnerability | Electrode mesh implanted in craniotomy, harmful to health |
Intracortical point signal acquisition | Electrical signals | ~0.003 s | ~0.1 mm (MUA) | Invasive | High spatial and temporal resolution | Signal quality and sensitivity diminish with time |
Functional MRI | Metabolism | ~1 s | ~1 mm | Noninvasive | High spatial resolution | Very low time resolution, too large to carry |
Near-infrared spectroscopy | Metabolism | ~1 s | ~5 mm | Noninvasive | Low cost, high portability, and acceptable time resolution on the order of 100 milliseconds | Very low spatial resolution |
The BCI operation’s signal processing phase is split into two sections. Feature extraction is the first step, which extracts the signal features that encode the user’s purpose. Different types of thought generate different patterns of brain signals, and the BCI classifies each pattern into a category based on their characteristics. Depending on the type of control signals in the BCI, the patterns can be divided into visual evoked potentials (VEPs), slow cortical potentials (SCPs), P300 evoked potentials (P300), or sensorimotor rhythms (
The second step in signal processing uses a conversion algorithm that converts extracted signal features into system commands. Brain electrophysiological features or parameters are converted into commands that will generate outputs, such as letter selection, cursor movement, and regulation of a motorized prosthetic, and influence additional assistive devices. The conversion algorithm must be dynamic to accommodate continuous changes in signal characteristics and ensure that the range of the user’s specific signal characteristics fully covers the range of control for the device [
End effectors are external output devices that are operated by BCI commands. The function and design of these devices vary depending on the intent of the BCI system and the needs of each end user. Here, we will focus on motor control and neurorehabilitation as end effector targets for BCIs, including functional electrical stimulation (FES), VR, intelligent wheelchairs, orthotics, and exoskeletal robotic devices.
Functional electrical stimulation (FES) technology works by sending electrical impulses to a paralyzed or damaged limb in order to produce artificial muscle contractions [
In recent years, BCI combined with VR technology has become a new technique that has significant applications in neurorehabilitation. Compared to traditional rehabilitation methods, BCI-VR systems can improve individual motivation by increasing the appeal of training, thus shortening the training cycle, providing more effective feedback, and facilitating recovery of brain function [
The main considerations of exoskeletal and orthotic devices for stroke patients are rehabilitation and replacement. An orthotic supports a joint as it moves from static to functional position and can also generate dynamic movements through the range of motion of the joint. This method is useful for patients with low motor neuronal disease or severe muscle atrophy. Ramos-Murguialday et al. developed a new combined EEG-EMG-based BCI technology for use in a BCI-operated hand orthotic neurorehabilitation system, and this technique has thoroughly demonstrated superiority over traditional cortical muscle coherence-based BCI classifications [
Robotic exoskeletons offer the advantages of increased joint strength and reduced load-bearing. Exoskeletons allow soldiers to lift heavy objects and can assist firefighters who have to wear heavy equipment. At the same time, exoskeletons can be utilized to assist the elderly or people with motor impairments in their daily activities. Combining these advantages with BCI technology, the exoskeleton-BCI system can enhance rehabilitation by providing the ability to repeat training exercises to increase the intensity of movement [
There is a desire to harness the potential of BCIs to transform the recovery process from neurological disease. Although it is still too early to apply the brain-computer end effector interface in a clinical rehabilitation setting, there has been significant progress in the integration process. Next, we will elaborate on BCI technology and its application in neurorehabilitation after stroke.
The use of BCI in stroke neurological rehabilitation is a new attempt in modern rehabilitation. Among them, the functionality of BCI in the remodeling of stroke patients’ central nervous systems is a pressing topic of inquiry. Neuroplasticity refers to the process by which the brain learns new behaviors, adapts to the environment, and modifies behavior by adding or changing existing synapses. The Hebbian theory, developed by Canadian cognitive psychologist Donald Hebb (1904-1985), suggests that repetitive stimulation of postsynaptic neurons by presynaptic neurons enhances the efficacy of synaptic transmission [
Current state of the application of BCIs in the field of stroke rehabilitation.
Current state of the development of BCIs in the field of stroke rehabilitation | |
---|---|
Motor rehabilitation | The use of BCIs is rapidly developing in the field of locomotion, and BCIs are effective in restoring upper and lower extremity motor when used in conjunction with FES, robotics, and robotic arms. |
Sensory rehabilitation | Related research is working on sensory-motor modalities for BCIs, and the development of sensory-motor closed-loop systems will improve the efficiency of rehabilitation. However, the development of sensory rehabilitation is still in its initial stage and has not yet been put into clinical use. |
Communication rehabilitation | BCIs can not only help restore the rehabilitation of language disorders in stroke but also serve as a substitute for language to restore the ability to communicate in patients with language loss. Currently, the study is based on three main signals: SCP-BCI, SMR-BCI, and P300-BCI. |
Cognitive rehabilitation | Applying BCIs to cognitive training improves certain cognitive functions in neurodevelopmental and neurodegenerative diseases, but there are relatively few clinical studies. |
Environment interaction | The application of BCIs in environmental interaction is the most humane consideration for the quality of life of stroke patients with hemiplegia. The development of smart homes is greatly increasing interactions between patients and the outside environment. |
Long-term motor disability as a result of stroke is one of the main targets of rehabilitation [
The restoration of upper limb motor defects in serious stroke patients was the initial impetus for the investigation of BCI technology in the field of poststroke rehabilitation. Patients with severe injuries do not have the minimum motor capacity required to undergo traditional rehabilitation therapies, such as occupational therapy (OT) or constraint-induced movement therapy (CIMT), which necessitates the search for a new kind of rehabilitation intervention [
Taylor et al. performed BCI-based interventions on healthy individuals and stroke patients and recorded motor-related cortical potentials by EEG during MI and ankle dorsiflexion in these subjects, demonstrating that BCI training can affect motor cortical excitability in the lower limbs of both healthy adults and stroke patients [
Sports rehabilitation author statistics.
Author | BCI facility | Site of action | Result |
---|---|---|---|
Ang | BCI-robotic | Upper limb | Improvement |
Pfurtscheller | BCI-FES | Hand | Improvement |
Caria | BCI-robotic | Arm and finger | Improvement |
Daly | BCI-FES | Hand and finger | Improvement |
Taylor | BCI-orthotics | Lower limbs | Improvement |
Chung | BCI-FES | Ankle | Improvement |
While BCIs have made great strides in motor control, significantly less attention has been paid to restoring tactile or skin sensation [
Speech production and communication comprehension deficits afflict up to 30% of people who have had a stroke [
People who have had a stroke also often suffer from cognitive impairment. Cognitive impairment can be seen as a range of deficits, including poor concentration, slowed information processing, memory impairment, reduced semantic fluency, difficulty producing or processing speech, and aphasia [
The effects of neurofeedback training based on BCIs have been shown to improve certain cognitive functions in neurodevelopmental and neurodegenerative disorders, such as attention-related hyperactivity disorder (ADHD) [
In the technical development of BCI-smart home control systems, the majority only consider younger healthier target populations. However, elderly people with disabilities or limited mobility are more interested in or in higher need of smart homes because of their limited capabilities; being able to operate household appliances at home on their own would greatly enhance their quality of life [
With the rapid development of modern technology, brain-computer interface (BCI) technology has become an extremely relevant topic in research, and the application of BCIs in the medical field has become one of the more important reasons encouraging its development. While this novel treatment for neurological rehabilitation after stroke is proving to be extremely beneficial, there are a number of areas that need improvement in the field of BCI research. The first area to address is the development of bidirectional BCIs. While we have currently made great progress in BCI motor control, we are at a distinct disadvantage for the recovery of tactile or cutaneous sensation, where the recovery of a limb requires the integration of both motor and sensory modalities. This is where a combination of a bidirectional BCI and a “closed-loop system” that integrates both motor output and sensory input is appealing, where the motor output is adjusted based on the sensory input to achieve the optimal motor route. A neuroprosthetic based on a bidirectional BCI is already under development, and it is just a matter of time until it is applied to stroke neurorehabilitation to better guide clinical practice. The second area that needs to be examined is the application of synchronous versus asynchronous BCI. The stroke BCI rehabilitation systems introduced in this paper, robotic arm, VR, and intelligent wheelchairs, are all synchronous BCI. A synchronous BCI system requires the user to set the EEG data acquisition experimental paradigm to obtain time-specific, real-time acquisition of data; therefore, the user is always “working” in sync with the system. In practice, however, users cannot be in a “working” state for long periods of time, and in most cases, they are usually in a “free” state. In order to improve the usability of online systems and avoid various errors, it is necessary to identify this idle state; to address this need, asynchronous BCIs have been created. There are studies on asynchronous BCI systems, but few of these regard applications to clinical rehabilitation. Asynchronous BCI movements would truly give patients full autonomy in their rehabilitation. Another exciting future possibility is the hybrid brain-machine interface, which requires the use of EEG as well as other physiological signals, such as neuromodulation (noninvasive brain stimulation), electromyographic activity, and heart rate, as input. These combinations can work synergistically to make the control algorithm more robust and to improve the reliability of the user’s intent to detect. For example, the existing BCI-robotic arm rehabilitation system only relies on EEG to drive the paralyzed limb in rehabilitation, and there is no active movement of the limb at all. However, under hybrid BCI control, we can use both EEG and EMG to jointly control the manipulator’s arm and enhance the “shared control” of the effector’s devices. Shared control between the preprogrammed control of the end effector device and the neural control of the human brain-computer interface has potential to improve the performance of motor tasks. Another phenomenon that must be explored is how the quality of signal acquisition of the BCI system directly determines the degree of execution of the effector. In terms of the current classification of signal acquisition methods, there are two types: invasive and noninvasive. Invasive BCI has the advantage of good signal quality, but the disadvantages are that it is very invasive and degrades over time due to damage to the electrode sheet. We therefore need to develop harmless, more stable electrode materials. The advantages of noninvasive BCI are its safety and convenience, but the disadvantage is its poor signal quality. We need to balance the advantages and disadvantages of both devices to find a more efficient way of collecting signals and to continue developing intelligent adaptive neural interfaces.
This review describes several BCI applications (e.g., motor, sensory, verbal, cognitive, and environmental interactions) to aid in the rehabilitation of stroke patients. We hope that the techniques presented in this paper will further contribute to the design of new applications and devices for BCI-based stroke rehabilitation. Brain-computer interface technology has already demonstrated exciting results in providing cognitive and physical support and rehabilitation, and we look forward to future innovations in this important area of research that will ultimately affect us all.
This article is a review article and does not contain relevant data.
The authors declare that there are no conflicts of interest regarding the publication of the paper.
Tiansong Yang and Xiaowei Sun contributed equally to this work. Siyu Yang and Ruobing Li are first authors.
This study was supported by grants from the National Nature Science Foundation of China (81503669, 81704170, and 82074539), Heilongjiang Province Natural Science Foundation (H2015031 and LH2020H092), outstanding Training Foundation of Heilongjiang University of Chinese Medicine (2019JC05), outstanding Innovative Talents Support Plan of Heilongjiang University of Chinese Medicine (2018RCD11), Heilongjiang Traditional Chinese Medicine Scientific Research Project (ZHY2020-125), and Postdoctoral Initiation Fund of Heilongjiang Province (LBH—Q18117).