The information is prepared on the basis of data from the information-analytical system RSF, informative part is represented in the author's edition. All rights belong to the authors, the use or reprinting of materials is permitted only with the prior consent of the authors.

Project titleDetection of quasi-movement sequences as a basis for an effective asynchronous brain-computer interface

Keywordsbrain-computer interface, EEG, asynchronous neural interface, human-machine interfaces, desynchronization, sensorimotor, quasi-movement, motor imagery, intention, kinesthetic imagery, pattern recognition, amplitude-frequency features

Annotation
Non-invasive brain-computer interfaces (BCIs) are increasingly being used in neurorehabilitation. More and more attempts are being made to develop BCI solutions for healthy users as well, particularly for cognitive training and simply for recreational purposes ("neurogames"). The most effective of the existing non-invasive BCIs are "synchronous", i.e., they require synchronization of the actions with the stimulus time set by the interface. This leads to a partial distraction of the user's cognitive resources to process these stimuli, as well as a decrease in the user's sense of control over the interface. Because the sense of control is a key component of the sense of agency (authorship of an action), the user of a synchronous interface may not feel themselves to be fully the author of an action. Both the diversion of cognitive resources to stimuli and the lack of full control when operating an interface reduce the usability of the interface. With this in mind, an "asynchronous" interface, in which the user can issue commands at arbitrary points in time determined by the user, appears to be the preferred type of BCI. However, the ability to issue a command at any point in time means that the BCI must check regularly and frequently whether the user has issued a command. Because non-invasive BCIs work with data (typically EEG data) in which the useful (for the BCI) signal is highly noisy, this results in either a high error rate or the need to accumulate large amounts of data to recognize the command -- in the latter case, interface response times increase significantly and the user may quickly become fatigued. Both low accuracy and long response times not only make working with a BCI difficult, but also, again, lead to a decreased sense of control and authorship (agency) of the action. This short-term (1 year) interdisciplinary project will evaluate the possibility of significantly improving the accuracy and speed performance of asynchronous BCIs by replacing the motor imagery traditionally used in them (that is, the user imagines the movements in his mind) by performing short sequences of quasi-movements at a fast tempo. Quasi-movements, a relatively recently discovered and poorly studied phenomenon, are observed when the subject is asked to gradually reduce the amplitude of a movement until both the movement and the electromyographic (EMG) signal from the corresponding muscles disappear. In this case, the pattern of changes typical of real movements is largely preserved in the EEG. Although the possibility of recognizing quasi-movements by the BCI classifier has already been studied and yielded positive results, in this analysis only synchronous BCIs were simulated. Moreover, quasi-movements, as observed by their first researchers, unlike imaginary movements, can easily be performed sequentially at a fast pace, which opens up the possibility of using their repetitions to obtain more clearly recognizable patterns of brain activation and, consequently, to improve the signal-to-noise ratio in the EEG and increase the accuracy of command recognition. The project will for the first time analyze the possibility of detecting a command using quasi-movements, and in particular sequences of quasi-movements, in the asynchronous mode of application of the BCI classifier. Such a mode will be simulated in the offline mode using data obtained in experiments with healthy participants performing quasi-movements and their sequences in response to stimuli, i.e. in the synchronous mode. The asynchronous mode would mean that the timing of the stimulus would not be used by its algorithm, which could be triggered at any time in a relatively wide time interval. By asking participants to perform quasi-movements in response to a stimulus with a known feed time, it will be possible to reliably estimate the delay in classifier response, the frequency of misses, and the frequency of false positives. A similar analysis will be applied to the data obtained in experiments with imagined movements and sequences of movements. Their conditions will be as close as possible to the conditions of experiments with quasi-movements, which will allow us to compare the time and accuracy characteristics of the BCI performance when using these two methods of command delivery without using real movements. If the hypothesis of the possibility of increasing the speed or accuracy of a BCI when using quasi-motions or their sequences compared to the use of imaginary movements and their sequences is confirmed, a new project proposal will be prepared that will implement an asynchronous real-time, quasi-movement-based BCI and a technique to test the hypothesis of increased sense of agency and control when using it compared to using a traditional motor imagery BCI.

Expected results
1. The methodology of experimental EEG-study of short sequences of quasi-movements will be developed and fine-tuned. 2. For the first time the effect of performing quasi-movements by short sequences in short (not more than 1 s) time intervals on EEG patterns accompanying them and features of these effects in comparison with real and imaginary movements will be described. 3. For the first time, we will obtain (in offline simulations) estimates of the accuracy and speed characteristics of asynchronous BCI based on quasi-movements and sequences of quasi-movements, and compare them with the characteristics of BCI based on imaginary movements. These results will make it possible to draw conclusions about the prospects of developing a new technology of BCI (brain-computer interfaces), which has the potential to become one of the leading technologies among asynchronous noninvasive BCIs.

Annotation of the results obtained in 2022
In the project, a new version of the methodology for quasi-movements practice and their comparison with imaginary movements was developed and refined; in particular, we developed a special training to distinguish between quasi-movements and imaginary movements. An experimental study was conducted in 23 participants, who participated in two sessions held on different days. In the main session, the EMG and 128-channel EEG were recorded, and the answers of the participants to the questions of specially designed surveys were collected. The methodology for analyzing EMG and EEG collected during quasi-movements and under control conditions has been significantly improved. In particular, it included individual selection of frequency ranges and spatial filters, which were based, respectively, on “superlets” (a modern variation of wavelet analysis with varying the number of cycles) and on the generalized eigendecomposition (GED). Statistical analysis was carried out primarily using mixed linear models, which made it possible to effectively separate the contribution of various factors that were significant in this study. Particular attention was paid to the precise assessment of microbursts of muscle activity. To simulate asynchronous classification, the recently developed well-interpreted compact convolutional artificial neural network SimpleNet (Petrosyan et al., 2022) was chosen as a classifier. The most important results of the project include the following: - we confirmed the observation of Nikulin et al. (2008) that quasi-movements are accompanied by a deeper suppression of sensorimotor rhythms than imaginary movements (contrary to the observations of Zich et al., 2015, who possibly did not train their participants to perform quasi-movements to sufficient extent) - for the first time, the independence of the suppression of cortical sensorimotor rhythms during quasi-movements from residual muscle activation was established (which is important for the development of methods for using quasi-movements in fundamental research and in the development of neurorehabilitation techniques) These results indicate that quasi-movements are not part of a continuum between real and imagined movements, as suggested by previous researchers, but are akin to attempted movements by people who are unable to perform real movements due to paralysis or amputation. This opens up the possibility of using quasi-movements to study the nature of performing actions, and also confirms the assumptions of Nikulin et al. (2008) that quasi-movements can be useful for modeling attempted movements, i.e., attempts to make movements by paralyzed patients and amputees, which is important for the development of neurorehabilitation techniques and brain-computer interfaces. Some of the publications on the project by the time of submission of the report were under review and in the process of finalizing the preparation of the manuscript. Links to these articles as they are published, as well as preprints, will be placed in the Projects section of the site of our research group ( https://bci.megmoscow.ru/projects/ ).

Publications

1. Shishkin S.L. Active Brain-Computer Interfacing for Healthy Users Frontiers in Neuroscience, 16:859887 (year - 2022) https://doi.org/10.3389/fnins.2022.859887

2. Vasilyev A.N., Yashin A.S., Shishkin S.L. Quasi-Movements and “Quasi-Quasi-Movements”: Does Residual Muscle Activation Matter? Life, Life 2023, 13(2), 303; https://doi.org/10.3390/life13020303 (year - 2023) https://doi.org/10.3390/life13020303

3. Yashin A.S., Shishkin S.L., Vasilyev A.N. Is there a continuum of agentive awareness across physical and mental actions? The case of quasi-movements Consciousness and Cognition, Consciousness and Cognition. — 2023. — Vol. 112. — P. 103531. (year - 2023) https://doi.org/10.1016/j.concog.2023.103531

4. Yashin A.S., Vasilyev A.N., Shishkin S.L. Contrasting quasi-movements with imaginary movements: an experimental model for studying physical and mental actions Fourth International Conference Neurotechnologies and Neurointerfaces (CNN), pp. 215-218 (year - 2022) https://doi.org/10.1109/CNN56452.2022.9912508

5. - VR, вождение и искусство: как нейроинтерфейсы входят в нашу жизнь РБК Тренды, 05.05.2022 (year - )

6. - «Активные» нейроинтерфейсы для здоровых людей: ближайшие перспективы Нейроновости, 25.04.2022 (year - )