MINI REVIEW: THE APPLICATION OF BRAIN-COMPUTER INTER- FACES IN ROBOTIC THERAPY

The idea of robotic therapy has been considered as a possible rehabilitation strategy to facilitate recovery of the patients with disability and it can represent an ef icient treatment. Brain-computer interface (BCI) is known as an advanced technology with great potential in therapeutic and assistive robots. This paper is presented to review the application of BCI in rehabilitation robotic systems through the combination of BCI with electroencephalography (EEG) and functional electrical stimulation (FES). For this purpose, the basic concept of each of BCI, EEG, and FES is introduced to give a general view of their function. In addition, the application of EEG-BCI and FES-BCI systems in therapeutic and assistive treatments is showed by providing a summary of different researches for each ield. In the end, this document is terminated with a discussion about the arguments behind the studied topics and the future directions of advances in robotic therapy.


INTRODUCTION
Stroke is a signi icant reason for acquired disability in the world, which can end with substantial neural damage or even death. It is a major reason for long-term disabilities and may lead to loss of movement, coordination, sensation, and dif iculties with activities of daily living Chung (2017); Louie and Eng (2016); Mozaffarian et al. (2016); Rodgers et al. (2019). Strokes in the elderly people may reduce their mobility and lead them to fewer independent lives and poor quality of life; even though they normally have physical deterioration and weakness, which imposes a heavy responsibility on the social health care system B. ; H. Kim et al. (2019). Also, millions of people experience stroke worldwide each year, and many of these patients have to deal with some level of permanent hemiparesis resulting from the damage to neural tissues Harcum et al. (2019); H. Kim et al. (2019). These stroke survivors and the elderly are beginning recognition of the importance of rehabilitation methods as a lifelong practice because they cannot do daily activities independently. However, traditional manual therapy methods mostly depend on the experience of the therapist and they often do not meet the requirements of high-intensity and repetitive training Zhou et al. (2013).
Robotic therapy is an advanced technique of physical therapy, through which patients practice their paretic limb by resorting to or resisting the force offered by a robotic system to enable the implementation of highly repetitive, intensive, adaptive, and quanti iable physical training Basteris et al. (2014); J. ; Duret et al. (2019). The application of rehabilitation robots can release the doctors and therapists from heavy training tasks and also assess the patient's functional performance by measuring kinetic movement parameters. Due to the bene its of its accuracy and reliability, rehabilitation robots can provide an effective way to improve the result of stroke or postsurgical rehabilitation Simonetti et al. (2016); . Noticeably, robotic therapy can modify motor functional recovery with improved movement completion time, more suitable employment of joint muscle groups, smoothness of motion, and improved inter-joint coordination of joints de Sousa et al. (2018); Prange et al. (2015).
The robotic therapy equipment can be designed for recovery in the lower limb and/or upper limb. Lower limb rehabilitation robots can be used as assistive devices to allow individuals with complete spinal cord injury to walk Sale et al. (2012). These devices may include a mechanism for steps posture and weight alleviation controlling to assist patients in simulating healthy individual's footsteps and practicing leg muscles Díaz et al. (2011). Also, upper limb rehabilitation robots consist of methods to rehabilitate the hand, wrist, elbow, and shoulder G. Kim et al. (2017). Using these devices for sensorimotor training can improve motor control of the shoulder and elbow and upper limb functional outcomes Mehrholz et al. (2015); Veerbeek et al. (2017). Based on the mechanical structure rehabilitation robotic devices can be divided into end-effector and exoskeleton systems Duret et al. (2019); Yue et al. (2017); Zhang et al. (2017). An end-effector device is often external to the body of patients, and it can generate the required force to the end of the user's extremity to help or resist the motion Gopura et al. (2016). These systems ensure that the patient is restricted to the speci ied range of motions, by controlling the paths along which their joints can move and maneuver. Also, they can apply resistance to the patient in the same alignment as they are experiencing in a virtual reality simulation to provide accurate feedback data McConnell et al. (2017); Mehrholz et al. (2015). An exoskeleton device can be worn on the body of patients. The joint and links of the robot have a direct connection with the human joints and limbs, respectively Gopura et al. (2016). Also, the portability of the exoskeleton has made it a good choice for patients in the later period of stroke when they can practice and train themselves at home Bos et al. (2016). This paper is presented to review the application of brain-computer interface (BCI) in the development of rehabilitation robotic systems. Accordingly, at irst, a conceptual description of BCI is provided to be used in robotic therapy. Then, two methods, including electroencephalography and functional electrical stimulation, are introduced to review their basic functions. Also, a summary of examples of some related research studies in therapeutic and assistive methods are represented. In the end, the arguments behind the mentioned ields are discussed and a possible con iguration of future advanced technologies in robotic therapy is suggested.

METHODS
The advances in robotic technology had made it possible to form new rehabilitation systems, which can increase the ef iciency and develop the range of application of therapeutic devices. "Brain-computer interface" is one of these technologies.

BRAIN-COMPUTER INTERFACE
Brain-computer interface (BCI) is an advanced method that connects our natural brain with and an external device, providing a new communication channel for brain signals to control external devices without using the natural neuromuscular routes. BCIs are often directed at researching, mapping, assisting, augmenting, or repairing human cognitive or sensory-motor functions He et al. (2015); Krucoff et al. (2016). By detecting the brain's signals and interpreting them, BCI allows generating control commands for external devices. The detection of the brain signals is performed either by the recording of its electrical activity or the related magnetic activity Lange et al. (2016). A BCI-based system contains at least four main steps: (1) extracting signals from the nervous system, (2) decoding the signals to predict user intent, (3) generating an output to affect the subject's environment, and (4) providing a feedback system to help the user re ine the output Krucoff et al. (2016). Figure 1 shows how BCI operates in a system and its input and output signals.
The bene it of BCI is the independence from any remaining muscular functions, which means that muscle fatigue is irrelevant. So, it can be used to restore lost or impaired functions of patients severely disabled by various devastating neuromuscular disorders or patients with damaged nervous systems Lange et al. (2016); Lazarou et al. (2018). BCI can promote long-lasting recovery in the motor function of chronic stroke patients with severe myasthenia and represents a promising strategy in severe stroke neuro-rehabilitation Ramos-Murguialday et al. (2019). Combined with BCI, the signal of the brain evoked by the individuals' spontaneous motor imaginary can be extracted and classi ied to directly control the robot. With this collaboration, robotic therapy can discover an arti icial pathway to replace the normal motor pathway of humans Teo and Chew (2014).
BCIs have the potential to provide two key bene its to disabled users: an alternate means of communication, and the ability to independently move around in and inter- Figure 1 The functionof BCI in a system act with their environment He et al. (2015). The use of BCI in stroke rehabilitation can be divided into two main approaches: a monitoring mechanism, with recorded brain signals serving as feedback data for concentration levels of the therapeutic practice; and a control framework in which an arti icial actuator is driven at demand Maier et al. (2019).

ELECTROENCEPHALOGRAPHY
Non-invasive BCIs relate to methods that extract brain signals without surgical procedures. The experimental setup ranges from large and relatively expensive options, like magneto-encephalography (MEG), to lightweight and inexpensive methods, like electroencephalography (EEG). Due to its availability and ease of use, EEG is widely used in clinical stroke rehabilitation He et al. (2015); Yuan and He (2014). This method is performed by measuring the electric ield on the patient's scalp through electrodes. The signal of the patient is not enough to actuate the motion of the paretic limb, but it is strong enough for the data acquisition instruments to collect Lange et al. (2016). EEG-based BCI operates by establishing a closed-loop neural interface. In other words, a BCI system uses raw functional cortical activity generated by EEG and translates it into a classi ied device command Kreilinger et al. (2012); Soekadar et al. (2015). Also, signal input is ampli ied and processed by a regression model that extracts particular amplitude changes or features and accounts for signal noise Kasuga et al. (2015); Young et al. (2014). With real-time processing, the reduced representation of brain activity is effectively translated into an output or feedback modality, often one that allows the desired task to be performed more easily Remsik et al. (2016). Some features of EEG-based BCI systems are: • To monitor and control brain activity of patients to improve their neural representation • To facilitate arm mobilization with higher motor gain for paralyzed arms and hands • Movement intention decoding performance and motor recovery • To engage the patients with tasks and acquire cortical signal of their imagination of a movement By using EEG-based BCI, voluntary movement or motion attempt of the limb activates the primary sensorimotor section in the brain, which is characterized by speci ic brain rhythms over the hemisphere contralateral to the limb in use Ang et al. (2015). EEG signals can be controlled by BCI to aid the patient, either by bypassing a physical impairment in the ability to control the limb or by highlighting engagement with a movement Ron-Angevin et al. (2017). In Barsotti et al. (2015); Frisoli et al. (2012) , EEG was used as a data acquisition system to record activity during BCI-training to support tasks like preparation of movement, reaching, grasping, and releasing. This method was also used to study the effect of BCI control of passive motion for stroke patients while they were imagining a reach and grasp movement using the affected limb Lu et al. (2020); Pereira et al. (2018). EEG-based BCI was used in Chowdhury et al. (2019) to estimate the quality of engagement of a patient with tasks by measuring the relationship between the brain and the muscle signals.
In Spüler et al. (2018), the effects of decoding performance on movement intention in a BCI system was studied by varying the cortical source of activity and the EEG frequency band in stroke patients.

FUNCTIONAL ELECTRICAL STIMULATION
Functional electrical stimulation (FES) is a subtype of neuromuscular electrical stimulation, which uses low-energy electrical pulses to arti icially generate body movements. It can be used to generate muscle contraction in paralyzed limbs to produce functions such as grasping, walking, and standing Moineau et al. (2019). In FES, the electrode mainly operates as a conductor, sending electrical charge from a power supply over the tissue. When this applied voltage between the active electrode and a second electrode generates an electric ield, charge transfer occurs, which as a result, forces electrical charge to low (Ho et al. (2014) Ho et al., 2014. Then, this electrical current elicits a response in excitable cells including neurons. Cochlear implants to restore hearing, phrenic pacemakers that aid respiration, cardiac pacemakers to ensure cardiac function, and deep brain stimulation to control tremors due to Parkinson's disease are examples of applications of electrical stimulation systems (Marquez-Chin and Popovic (2020)). Nowadays, FES-BCI systems are increasingly being explored as potential rehabilitation tools for improving the function of partially impaired limbs (Do et al. (2011) ). FES can use real-time feedback of BCI signal input to optionally administer feedback responses of treatment only when the correct brain signals are discovered. Using FES in robotic therapy can lead to a better realization of the recovery of limb function (Rong et al. (2012) ). Some features of FES-BCI systems are: • To generate muscle contractions and subsequent limb movement • To provide functional recovery and elicit motor recovery • To adjust the intensity of the electrical stimuli to achieve more precise movements • To improve functional mobility and range of motion of lexion and extension muscles exercises Neuro-rehabilitation methods employing FES to maneuver the paralyzed muscles can postpone or prevent many secondary medical complications and modify functional independence by providing a means to exercise and practice physical obstacles Ho et al. (2014). These methods can be used to improve functional mobility and range of movement of paretic stroke patients T. Kim et al. (2016). FES-BCI training can be bene icial on shoulder reduction and active movements by facilitating motor recovery Jang et al. (2016). In Grimm et al. (2016), the effects of this method were examined on the range of motion and cortical modulation in stroke patients who performed wrist lexion and extension exercises. BCI-FES can provide high functional recovery in patients with a high degree of residual mobility in arm and hand function and also elicit motor recovery in stroke patients Biasiucci et al. (2018). To perform daily living activities, BCI with a neural decoder can be used to produce coordinated reaching andgrasping movements and FES to generate muscle contractions and subsequent limb movement Ajiboye et al. (2017). In Bockbrader et al. (2019), the ability of FES-BCI was examined in evoking greater wrist extension strength and controlling translation, orientation, and hand shape of robotic limbs.

DISCUSSION
Robotic therapy is known as an ef icient rehabilitation technology to facilitate the recovery and treatment of patients with disabilities. BCI is an advanced technology with great potential in rehabilitation robots that can control external devices using brain signals. Accordingly, as it was reviewed in this paper, it can be de ined a control loop for rehabilitation robots including BCI, EEG, and FES. In this loop, EEG, as a noninvasive method, provides brain signals for BCI. Then, BCI processes the acquired signals and controls the robotic device. Also, FES can be used as a part of this device to facilitate the movements of the limb. In addition, there are some methods, like virtual reality and augmented reality, which can provide visual and auditory feedback signals to be used in this loop. It is estimated that by the collaboration of BCI, EEG, and FES with a rehabilitation robotic system, even the patients with more serious disabilities can bene it an effective treatment.