Brain-Computer Interfaces in Disorders of Consciousness

The development of brain-computer interfaces (BCIs) has established a new communication channel between the brain and external devices for information transmission that requires no muscular signals [1]. BCIs have been preliminarily studied to improve motor functions in patients with severe motor disabilities, especially lock-in syndrome. At present, the application of BCIs has been extensively validated. An emerging field of such research involves patients with disorders of consciousness (DoCs), in which BCIs can act as diagnostic, communication, and rehabilitation tools. Patients with DoCs have difficulty perceiving themselves and the surrounding environment after severe brain injury. DoCs include 3 states: coma, vegetative state (VS, also called unresponsive wakefulness syndrome), and minimally-conscious state (MCS). In patients with DoCs, there are inconsistent signs of arousal and awareness, which lead to misdiagnosis on traditional behavioral scales, such as the Coma Recovery Scale-Revised (CRS-R) [2]. In recent years, electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) have been used to detect residual brain functions in certain patients with DoCs [3, 4]. Naci et al. compared the application of fMRI, functional near-infrared spectroscopy, and EEG signals in BCIs [5]. The study pointed out that EEG signals have relatively low spatial resolution (especially for deep brain structures), they are vulnerable to muscle and eye movement artifacts, and the existing paradigm is difficult to apply to patients with DoCs [5]. Nevertheless, EEG is still the most promising technology due to its simplicity, low cost, high temporal resolution, lack of physical limitations, and longstanding history of application. In this paper, we aim to assess the challenges and possibilities for applications that could improve the quality of life in patients with DoCs.

Diagnosis

In the application of BCIs in patients with DoCs, the primary goal is to screen specific components of brain signals that are relevant to command-following as evidence of the presence of consciousness [6].

Owen et al. first studied the possibility of BCI application in detecting command-following: the paradigm of motor imagination [7]. The results showed that a patient in a VS was able to perform motor imagination (playing tennis and navigating through the room) as required in the fMRI test. P300 is a component of event-related potentials; it is a late positive wave with a latency of ~300 ms evoked by stimulation, and is now widely used in EEG assessments. Recently, Pan et al. proposed an audio-visual hybrid BCI based on the auditory P300 and steady-state visual evoked potentials [8]. The results showed that 4 healthy subjects, 1 patient with VS, 1 patient with MCS, and 1 patient with locked-in syndrome (LIS) exhibited sufficient command-following. Therefore, a BCI was considered to be a supportive bedside tool for evaluating consciousness in patients with DoCs. Some studies have also shown that the presence of P300 can be used as a prognostic indicator. Schnakers et al. proposed the detection of command-following using auditory P300 [9]. Therefore, BCIs are able to serve as supplementary behavioral assessment tools for the CRS-R in the evaluation of consciousness, since it better distinguishes patients with VS from patients with MCS.

Communication

The effectiveness of EEG-based BCI systems in communication has been preliminarily verified in volunteers and patients with amyotrophic lateral sclerosis (ALS) [10]. Among the patients with DoC, Li et al. used a hybrid BCI paradigm based on P300 and steady-state visual evoked potentials in 6 patients with VS, 3 with MCS, and 2 who emerged from MCS [11]. The results revealed that cognitive communication can occur in patients with DoCs using a BCI system.

Several studies have used auditory P300 BCIs in patients with DoCs [12]. In this paradigm, an enhanced P300, which can be considered to represent command-following, was found in some patients with MCS. However, the latency of P300, which is associated with information processing, is prolonged in patients with MCS [13]. Lulé et al. used an auditory P300-based BCI to test following commands in 18 patients [14]. These patients and controls were presented 4 auditory stimuli (‘‘yes’’, ‘‘no’’, ‘‘stop’’, or ‘‘go’’) after listening to a closed question, and then they were instructed to concentrate on either yes or no depending on the answer to the question. The results showed that P300 appeared in almost all healthy participants after stimulation, 1 patient with LIS, and 1 patient in a MCS, while P300 was not detected in the other patients with DoCs after stimulation. In practice, the auditory paradigm is more suitable for patients with eye-movement dysfunctions or visual abnormalities, while the accuracy rate of the auditory paradigm versus visual paradigm need further research.

To combine the advantages of audio and visual stimulation, Wang et al. designed a paradigm with audio-visual stimulation for 7 DoC patients and asked them to choose the corresponding numbers according to instructions [15]. Byrne et al. reported that a patient in a VS correctly answered 5 simple questions about his family's name with the help of a BCI system [16].

In addition to visual or auditory paradigms, Guger et al. used the tactile P300-based BCI to assess the command-following and communication abilities of patients with DoCs [17]. Patients were required to mentally count the target vibration times, and the EEG signals were recorded from 100 ms before to 600 ms after stimulation. In future, EEG-based BCI systems still play a vital role in research on communication in patients with DoCs.

When imagining limb movement, the mu rhythm [8–12 Hz, sensorimotor rhythm (SMR)] and beta band (13–30 Hz) in the motor area change significantly. When preparing to move or during movement, the marked decrease in the SMR in the contralateral motor cortex is commonly referred to as event-related desynchronization, and the significant ipsilateral increase in the SMR occurs at the end of a movement, which is commonly referred to as event-related synchronization. Benzy et al. showed that rhythms based on motor imagination can be used as the control signal of BCIs [18]. The results showed that 19% of patients with MCS or VS have sensory and motor rhythm activation, and some patients show continuous attention, response choices, working memory, and language comprehension. Cruse et al. also studied the paradigm of motor imagination in the diagnosis of VS and showed a diagnostic accuracy of 53% [19].

However, consistent performance is difficult to achieve in the motor imagination paradigm due to the need for long-term training of motor imagery-based BCIs and the impact of volitional fluctuations. In addition, although the recognition rate of two types of EEG signal of motor imagery is high, it is still a challenge to differentiate three or more types of EEG signal in response to motor imagery. Therefore, research is still needed on motor imagination-based BCIs in the field of DoCs.

Rehabilitation

In essence, a DoC is the loss of function in certain eloquent areas, and the remaining functional areas lack sufficient connection or integration to support arousal and awareness. Based on this definition, it is believed that the recovery of consciousness depends, to a certain extent, on remodeling of the central nervous system [20]. With regard to electrical nerve stimulation, it is generally accepted that cervical spinal stimulation at the C2–C5 levels or deep brain stimulation of crucial central nuclei both improve the condition of patients, especially those whose DoC is due to traumatic brain injury. BCI feedback-based electrical stimulation can be used to improve the condition of patients, especially after trauma. Closed-loop deep brain stimulation and spinal cord stimulation are the main research directions of BCIs for rehabilitation of DoCs [20]. Studies based on neural circuits are needed to explore the mechanisms. However, for a special population, such as patients with DoCs, a specific BCI rehabilitation system appropriate for their condition has not been well developed. Due to the complex rehabilitation mechanisms after brain injury, further clinical verification is needed to determine whether multimodal BCI training can promote neural remodeling and more quickly improve the prognosis for patients with DoCs.

Challenges

Although many studies have shown that BCIs have been successfully applied to patients with DoCs, the high false-negative rate of BCIs cannot be ignored. When designing a BCI for the diagnosis of patients with a DoC, the use of more accurate techniques for judging the patient's state of consciousness should be considered. At the same time, since patients may be in the same state for a long period of time, the BCI for patients with a DoC should be able to obtain stable results in the long term to avoid fluctuations among repeated measurements. In the monitoring of patients with a DoC, many are taken care of by their families most of the time, so in routine outpatient reviewing, BCIs that realize automated and portable measurement may be necessary. In fact, patients with DoCs, especially in an MCS, often suffer from awakening fluctuation, fatigue, and difficulty in concentrating attention. Therefore, the complexity of the experimental paradigm (stimulus and instruction), duration, and repeatability of test results are important factors to be considered when applying BCI systems for diagnosis and evaluation. Meanwhile, it is difficult for patients with VS to communicate or give answers, and a BCI paradigm requiring no training may be more suitable for these patients. Some patients with MCS may cooperate with command following, and others may show delirium. In this case, higher requirements are placed on maintaining the stability of the test results, and the interference of body movements on the paradigm should be avoided as much as possible. In addition, brain injury often causes sensorimotor dysfunctions, such as cortical deafness, blindness, and eye movement disorder. Most patients with DoCs cannot concentrate for a long time, which may lead to negative BCI classification results. Therefore, when developing a BCI system, standard forms are required for patients with various sensory defects to reduce the number of false-negative results. At present, there is little research on algorithms of BCIs for patients with DoCs. The research of Jin et al. may help to enhance the signal-to-noise ratio of BCI, thereby improving the accuracy of detection for patients with DoCs [21–23]. Study has also shown that machine learning-based BCI may be helpful in diagnosis and this needs further research [24]. It is particularly important to design a reasonable experimental paradigm, keeping in mind that a successful experimental paradigm partially depends on the patient's willingness to complete the task. Patient willingness further decreases in the case of loss of motivation or immobility mutism. However, the above factors must be carefully considered. It is currently technically impossible to distinguish between patients with DoCs who lack motivation to complete the task and those who are not aware at all. In the interpretation of BCI results, negative results should be carefully interpreted, as the possibility of non-detection by patients with implicit consciousness is significantly different in different experimental paradigms. Similarly, positive results should not be regarded as clear evidence of the existence of consciousness but rather as an opportunity for better clinical outcomes [25] (Table 1).

Table 1.

BCI studies in patients with DoCs.

Study, year	Signals/EEG-based paradigms	Patients	Samples	Measure (test)	Invasiveness	Diagnostic accuracy/Rehabilitative efficiency	Diagnostic or rehabilitative	Patient training required (yes/no)
Classen et al. [24], 2019	Machine learning EEG	104 acute DoC	104	GOS-E	Non-invasive	–	Diagnostic	Yes
Hermann et al. [26], 2021	FDG-PET and audio-P300	21 VS/UWS, 31 MCS	84	CRS-R	Minimal-invasive	85%	Diagnostic	No
Cruse et al. [27], 2011	Motor imagination EEG	16 VS/UWS	28	CRS-R	Non-invasive	53%	Diagnostic	Yes
Guger et al. [17], 2018	Vibrotactile-P300	12 VS/UWS	15	CRS-R	Non-invasive	26%-38%	Diagnostic	Yes
Li et al. [11], 2015	SSVEP-P300	6 VS/UWS, 3 MCS, 2 eMCS	15	CRS-R	Non-invasive	61%	Diagnostic	No
Lulé et al.[14], 2013	Audio-P300	3 VS/UWS, 13 MCS, 2 LIS	34	CRS-R	Non-invasive	–	Diagnostic	Yes
Monti et al. [16], 2010	fMRI	23 VS/UWS, 31 MCS	70	CRS-R	Non-invasive	–	Diagnostic	No
Pan et al. [8], 2014	SSVEP-P300	4 VS/UWS, 3 MCS, 1 LIS	12	CRS-R	Non-invasive	69%	Diagnostic	No
Pan et al. [13], 2018	SSVEP-P300	5 VS/UWS, 3 MCS	16	CRS-R	Non-invasive	57%	Diagnostic	Yes
Schettini et al. [12], 2015	Audio-P300	5 VS/UWS, 8 MCS	25	CRS-R	Non-invasive	–	Diagnostic	No
Xiao et al. [28], 2018	Visual-P300	8 VS/UWS, 5 MCS, 1 eMCS, 1 LIS	20	CRS-R	Non-invasive	35%	Diagnostic	No
Schnakers et al. [9], 2008	Audio-P300	8 VS/UWS, 14 MCS	22	CRS-R	Non-invasive	–	Diagnostic	No
Wang et al. [15], 2015	Audiovisual-P300	3 VS/UWS, 4 MCS	17	CRS-R	Non-invasive	66%	Diagnostic	Yes

DoCs, disorders of consciousness; GOS-E, Glasgow Outcome Scale-Extended; VS/UWS, vegetative state/unresponsive wakefulness syndrome; SSVEP, steady-state visual evoked potential; MCS, minimally conscious state; eMCS, emerged from MCS; LIS, locked-in syndrome; CRS-R, Coma Recovery Scale-Revised

At present, research on BCI applications in patients with DoCs is a promising direction, but many problems remain to be solved. Further research is needed to address the current problems. The keystone of BCIs used in patients with DoCs is the detection of brain responses to external stimuli. Through the above comparison, the application of audio-visual multimodal P300 is relatively more promising. However, due to the fluctuation of consciousness in patients with DoCs and the short time period during which patients focus their attention, it is difficult to achieve performance consistency in the BCI paradigm. Currently, the complex paradigm takes an extensive amount of time for training; moreover, it is still difficult for clinicians to standardize the protocol, which hinders its promotion and application in the clinic. Looking to the future, the paradigm of simplifying BCIs will be convenient for most clinicians to operate and shorten the training and testing time. In this way, BCIs will be more suitable for clinical use in patients with DoCs. With the help of BCIs, the misdiagnosis of patients with DoCs can be reduced; thus, a new rehabilitation technology can be developed that enables personalized treatments for patients with DoCs.