The brain-computer interface (BCI) constructs a direct information transmission path between the brain and the external world by decoding the information of brain neural activity during human thinking activities. In the fields of neural prosthesis, neurofeedback training, brain state monitoring, etc. Wide application prospects. 2017 can be said to be a new starting point for the development of brain-computer interface. Under the impetus of a series of scientific research breakthroughs in brain-computer interface in recent years, brain-computer interface technology has begun to mature, and more and more attention and input from the industry is being obtained. For example, Elon Musk, a leader in entrepreneurship, has invested in the creation of Neuralink, a brain-computer interface company for neuroprosthetic applications and future human-computer communications. Facebook, the Internet leader, has announced the development of a new generation of interactive technologies based on brain-computer interfaces. These actions have caused heated discussions and will further accelerate the development of brain-computer interface technology. Based on the research progress of brain-computer interface in 2017, this paper first introduces the important achievements of the brain-computer interface in the implementation of the application system, then introduces the new progress in the application of key technology research, and finally forecasts the future development trend.
First, the application system to achieve: more effective communication and control
Character input
One of the most important application goals of the brain-computer interface is to enable patients with severe movement disorders to regain communication with the outside world. Following the breakthrough in the research of character input based on scalp EEG in recent years, in February 2017, the research team at Stanford University in the United States reported a high-performance brain-computer interface application system that uses intracranial EEG for character input. In this system, the investigators implanted high-density microelectrode arrays in the region of the motor cortex responsible for hand movement in patients with lateral sclerosis and spinal cord injury to acquire action potentials and high-frequency local field potentials, decoding these electrodes. The neural activity information enables continuous control of the two-dimensional cursor on the screen and a "click" action of character selection, thereby allowing the patient to interact with the outside world by inputting text through the on-screen virtual keyboard (Fig. 1). Three patients who used the system achieved 39.2, 31.6, and 13.5 English characters per minute, which is the fastest information transmission rate currently achieved in patients with dyskinesia. Compared with scalp brain electricity, the brain electrical signal obtained by implanting the intracranial electrode has higher signal-to-noise ratio and more stable signal, and has unique advantages in clinical application of the brain-computer interface for paralyzed patients. The study was published in the frontier journal eLife in the field of neuroscience and neuroengineering.
Figure 1 A brain lateral sclerosis patient enters characters through a brain-computer interface system
Motion control
Controlling the mobility aids such as wheelchairs and robotic arms is as important as character input for improving the quality of life of patients with severe dyskinesia. It is also the key research direction in the field of brain-computer interface. In May 2017, research from the Case Western Reserve University research team in the world's leading clinical medical journal The Lancet went one step further in this direction. They combine brain-computer interface with functional electrical stimulation technology to allow patients to control their own limbs to interact with the outside world.
Functional electrical stimulation (FES) is a technique that applies electrical stimulation to peripheral nerves and muscle tissue to enable high paraplegic patients to regain limb mobility (Fig. 2(a)). In this study, the researchers used the implantable brain-computer interface system to directly extract the cranial nerve signals in the hand-sports area of ​​the motor cortex in patients with traumatic high-grade cervical spinal cord injury. The spectral energy and action potential of the high-frequency band exceeded the threshold. The number of times is characterized by decoding the stimulation parameters of the corresponding functional electrical stimulator, controlling the functional electrical stimulator, and outputting electrical stimulation to the patient's squatted arm to induce muscle activity. After a certain period of training, the patient can achieve the reach and grab action. In the test phase, the patient successfully completed the use of his own arm and palm in 11 of the 12 attempts. The coffee mission (Fig. 2(b)), each task time is 20~40 s. The brain-computer interface system can help patients achieve continuous motion control, close to the actual life mode of operation, and is expected to help patients achieve natural smooth motion control of the limbs in the future, thereby greatly improving the quality of life of patients.
Figure 2 combined with intracranial brain-computer interface and functional electrical stimulation to achieve closed-loop control of the affected limb
The motion control application based on the brain-computer interface can not only serve the user's control needs, but also have more medical significance. Phantom limb pain is a disease in which patients feel pain in their limbs that are not present or have been paralyzed. There is currently no effective treatment. The research team at Osaka University in Japan found that using brain-computer interface training to weaken the representation of the phantom limb in the sensory motor cortex can effectively reduce the pain of the phantom limb. Brain-computer interface training is expected to become a clinically visceral pain treatment. . The research was published in the journal Nature Communications and won the 3rd place in the 2017 Annual BCI Award.
Completely blocked state patient information output
Patients with lateral sclerosis are an important target patient population for the brain-computer interface system. Lateral sclerosis, also known as gradual freezing, develops to the most severe condition, the patient will completely lose control of all muscles, this condition is called completely locked-in state (CLIS). Existing brain-computer interface systems often rely in part on the patient's residual nerve and muscle control capabilities and are not suitable for such patients. In January 2017, the research team from the University of Tübingen, Germany, based on functional near-infrared brain imaging, achieved the first brain-computer interface system for patients with CLIS. The researchers classified the changes in blood oxygen content in the middle part of the frontal lobe collected by the near-infrared function. After training for several weeks, the four patients could answer a series of speech problems by adjusting the brain neural activity pattern to answer “yes†or "No", the correct rate is around 70%. The work was published in the journal PLoS Biology.
Second, the application of key technology advances: new hardware, new algorithms, new paradigms
New hardware
The construction of the brain-computer interface practical system puts high demands on the miniaturization and wirelessization of the brain nerve signal acquisition and processing equipment. Although many small-scale devices based on EEG have been introduced, miniaturized hardware that can simultaneously collect a variety of neurophysiological signals is rare. In June 2017, the brain-computer interface research team at the Technical University of Berlin in Germany released a wireless device that can simultaneously collect EEG, functional near-infrared brain function images, and other conventional physiological parameters (such as ECG, EMG, and acceleration). M3BA: a mobile, modular, multimodal biosignal acquisition architecture, where each M3BA module (without battery) has a side length of only 42 mm (Figure 3). This is the first multi-physiological parameter acquisition architecture that combines modularity, mobilization, miniaturization, multi-modality and scalability. It has positive significance for realizing the practical application of brain-computer interface technology.
Figure 3 M3BA concept map
Signal acquisition is an important factor that restricts the brain-computer interface from the laboratory to real-life applications. For the non-invasive brain-computer interface based on scalp EEG, the brain-electrode interface system of the wet electrode solution takes a long time to prepare the conductive paste, and needs to clean the hair after use; while the conventional dry electrode system mainly collects the EEG signal of the forehead area. The resulting decoded brain has few states and low accuracy, and the actual use scenario is limited. The research team at the University of California, San Diego, in "IEEE Transactions on Neural Systems and Rehabilitation Engineering," reported a steady-state visual evoked potential brain-computer interface system that placed electrodes behind the ear without hair coverage, and in a 12-category task. In the middle, the classification accuracy rate is about 85%, and the information transmission rate is about 30 bit/min (Fig. 4(a)). This research progress provides a feasible support for the common healthy people to collect EEG in daily life situations and use the brain-computer interface efficiently. Correspondingly, the brain-computer interface commercial hardware based on the back of the ear is now appearing (Fig. 4(b)), which is expected to push related applications to the market quickly.
Figure 4 A new attempt at the post-ear brain acquisition site
New algorithm
How to improve the information transmission rate has always been an important topic in the field of brain-computer interface research. Limited by factors such as low signal-to-noise ratio of brain signals, the information transmission rate of brain-computer interface is lower than that of normal output channels (ie, peripheral nerve and muscle tissue), such as the classic P300 brain-computer interface character spelling system. The information transmission rate is approximately 0.5 bit/s. In 2015, Tsinghua University and the cooperative research team reported a high-speed brain-computer interface system based on steady-state visual evoked potentials, achieving a communication rate of 4.5 bit/s. In 2017, the Institute of Semiconductors and the cooperative research team of the Chinese Academy of Sciences made further breakthroughs and proposed a task-related component analysis algorithm to further increase the communication rate of the steady-state visual evoked potential brain-computer interface to 5.4 bit/s, and the optimal result reached 6.3 bit/ s, is the fastest scalp brain computer interface system that has been reported so far, and is expected to promote the application of brain-computer interface in the daily life of ordinary healthy people. The work was published in IEEE Transactions on Biomedical Engineering.
In addition, the research team at the University of Aalborg in Denmark proposed an online brain-computer interface algorithm that can monitor changes in the subject's attention and perform adaptive feature extraction in real time. The algorithm still maintains good performance in long-term use. This achievement was awarded the 1st place in the 2017 Annual BCI Award.
New paradigm
The brain-computer interface based on visual evoked potential is the type of brain-computer interface with the highest information transmission rate. Currently, in this type of system, each instruction corresponds to a specific visual code (time, frequency or pseudo-random coding, etc.). The specific electroencephalographic response induced by the coding is used to achieve target recognition. In 2017, the research team at the University of Hamburg in Germany proposed a new paradigm for the steady-state visual evoked potential brain-computer interface based on spatial information coding, using only one steady-state visual stimuli to achieve the identification of multiple attention targets. The new paradigm is based on the visual cortex retinal mapping principle. When the user pays attention to the different spatial orientations of the visual stimulus, the spatial pattern of the induced EEG response is different, and the correct classification rate of the offline 9-direction is about 95%. Subsequently, the University of Hamburg collaborated with Tsinghua University to implement a four-category online system that achieved around 90% of online classification performance in an actual two-dimensional motion control game. In such a paradigm, since there is no need to look directly at the steady-state visual stimuli, the user's visual load is lower and the user experience is better. At the same time, the design is more economical to use the computer screen, and can be better integrated in the more natural. Or a complex application background, there is a good practical prospect.
Imaginative movement is another major paradigm type commonly used in brain-computer interfaces, but there is no standardized solution for how to perform effective imaginative motion task training. The research team of East China University of Science and Technology has proposed that the user imagines that writing Chinese characters by hand as an imaginary movement task paradigm has achieved a classification performance that is significantly higher than the traditional paradigm. The paradigm carries out paradigm task design for the characteristics of domestic user groups, and provides a good idea for the promotion and application of brain-computer interfaces in China.
Third, development trends and prospects
Machine learning algorithms and data normalization
A new generation of machine learning algorithms represented by deep learning is gaining more and more attention from researchers in the field of brain-computer interface, which is expected to reduce the workload of researchers in extracting neural data features while maintaining excellent classification performance. It is worth mentioning that the research teams of Shanghai Jiaotong University, South China University of Technology, Xi'an Jiaotong University and other universities have achieved good results in brain-computer interface emotion recognition, P300 brain-computer interface attention target recognition, and imaginary motion classification.
Due to the relatively complex structure of the classifiers of the emerging machine learning algorithms, higher requirements are placed on the amount of data. In this context, several leading brain-computer interface research teams began to promote data standardization and open sharing: a joint research team between Tsinghua University and the Institute of Semiconductors of the Chinese Academy of Sciences released a set of 35 subjects with 40 stimulation frequencies. The standard data set for steady-state visual evoked potential brain-computer interface; the research team at the Technical University of Berlin in Germany introduced a hybrid brain-computer interface dataset that simultaneously collects EEG and near-infrared brain function image information, including data from 29 subjects. It includes classic brain-computer interface tasks such as imagination movement and mental arithmetic. The above data sets are available for free download by global researchers, and have positive significance for promoting algorithm research in the field of brain-computer interface.
Brain-computer interface and neuroethics
As the brain-computer interface technology gradually matures and approaches applications, the ethical issues that this technology may cause have begun to attract more and more attention and discussion. In November 2017, 25 well-known scholars in the field of brain-computer interface, neuroengineering and artificial intelligence jointly published a review article in the journal Nature, and proposed four major ethical issues in the development of neuroengineering technology. The first is privacy. Compared with other physiological parameters, neural signals carry more abundant personal information. How to protect the privacy of users needs to be considered in the development of related technologies. Secondly, identification is for long-term implantation. The brain stimulator group, how to determine whether its behavior is subject to its own control or device control, needs to be defined at both moral and legal levels; once again, it is to strengthen the problem, if using neuroengineering technology to create super agents who are stronger than ordinary people for war, Whether it violates social norms; finally, it is a prejudice problem. How to establish a relatively fair technical development norm that can take into account the interests of various groups. If there is some kind of prejudice rooted in neural equipment, it may cause serious social problems.
Opportunities and Challenges
2017 is a year of opportunity in the field of brain-computer interface. It is estimated that nearly 100 million US dollars are invested in brain-computer interfaces and other neuro-engineering fields every year, and this number is still growing. In addition to the industry concerns represented by Neuralink and Facebook, governments have also attached great importance to this technology. The brain plans of the United States, the European Union, Japan, and China will provide key neurophysiological foundations and key technical methods for brain-computer interfaces; and the Defense Advanced Research Projects Agency (DARPA) has released a design called Neural Engineering Systems Design. The new project proposes the goal of a wireless brain implant device that can simultaneously record brain neural signals with 1 million electrodes and selectively activate 100,000 neurons (Figure 5).
Figure 5 Conceptual view of a wireless brain implant device
There are also many challenges in the field of brain-computer interfaces. In October 2017, "IEEE Transactions on Neural Systems and Rehabilitation Engineering" published a special issue of IEEE Brain Initiative Special Issue on BMI/BCI Systems. The editorial review article summarizes several major issues that need to be addressed in the field of brain-computer interface. The question, including how the brain-computer interface system maintains stable performance in daily applications, how to design and implement multi-channel, low-power, long-life wireless EEG implant equipment, how to push the brain-computer interface to clinical practice. With increasing human and financial investment, brain-computer interface researchers are already actively preparing for the breakthrough of future theories and technologies.
Bluetooth Speaker,Outdoor Bluetooth Speaker,Hifi Bluetooth Speaker,Round Bluetooth Speaker
Comcn Electronics Limited , https://www.comencnspeaker.com