Artificial Intelligence Algorithms in Visual Evoked Potential-Based Brain-Computer Interfaces for Motor Rehabilitation Applications: Systematic Review and Future Directions

Abstract

Brain-Computer Interface (BCI) is a technology that uses electroencephalographic (EEG) signals to control external devices, such as Functional Electrical Stimulation (FES). Visual BCI paradigms based on P300 and Steady State Visually Evoked potentials (SSVEP) have shown high potential for clinical purposes. Numerous studies have been published on P300- and SSVEP-based non-invasive BCIs, but many of them present two shortcomings: (1) they are not aimed for motor rehabilitation applications, and (2) they do not report in detail the artificial intelligence (AI) methods used for classification, or their performance metrics. To address this gap, in this paper the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) methodology was applied to prepare a systematic literature review (SLR). Papers older than 10 years, repeated or not related to a motor rehabilitation application, were excluded. Of all the studies, 51.02% referred to theoretical analysis of classification algorithms. Of the remaining, 28.48% were for spelling, 12.73% for diverse applications (control of wheelchair or home appliances), and only 7.77% were focused on motor rehabilitation. After the inclusion and exclusion criteria were applied and quality screening was performed, 34 articles were selected. Of them, 26.47% used the P300 and 55.8% the SSVEP signal. Five applications categories were established: Rehabilitation Systems (17.64%), Virtual Reality environments (23.52%), FES (17.64%), Orthosis (29.41%), and Prosthesis (11.76%). Of all the works, only four performed tests with patients. The most reported machine learning (ML) algorithms used for classification were linear discriminant analysis (LDA) (48.64%) and support vector machine (16.21%), while only one study used a deep learning algorithm: a Convolutional Neural Network (CNN). The reported accuracy ranged from 38.02 to 100%, and the Information Transfer Rate from 1.55 to 49.25 bits per minute. While LDA is still the most used AI algorithm, CNN has shown promising results, but due to their high technical implementation requirements, many researchers do not justify its implementation as worthwile. To achieve quick and accurate online BCIs for motor rehabilitation applications, future works on SSVEP-, P300-based and hybrid BCIs should focus on optimizing the visual stimulation module and the training stage of ML and DL algorithms.

Identifying Thematics in a Brain‐Computer Interface Research

“…Te change in research focus was also evidenced through the shift in language use, with terms such as "speller" appearing in reviews after 2018 [87][88][89]. Other words that had a greater frequency of use in the later cloud were "classifcation," "data," "design," "information," and "learning" which are closely associated with refning BIC systems, and the associated growth with BCI, AI and DL [10][11][12][14][15][16][17][18], all of which appear in reviews post 2018. Te term "hybrid" was a word that became less frequently used in 2018.…”

“…As an enhanced understanding of the brain is achieved, there will a generation with further advances in BCI performance as task-specifc activity patterns and their ability to become accurately detected, their features comprehended, optimized, and classifed [7][8][9]. Tis intrinsically ties to the development of software systems linked to artifcial intelligence (AI), including machine learning (ML), pre-processing, and deep learning (DL) that will improve fexibility, extendibility, usability, and performance at the individual user level [10][11][12][13][14][15][16][17][18].…”

Non-invasive EEG-based BCI spellers from the beginning to today: a mini-review

“…Electroencephalography (EEG) indeed remains the most used BCI technique. BCI, in turn, is becoming an increasingly reliable method of experimental and clinical application due to the introduction of artificial intelligence (AI) into the brain-computer interface technology ( Gutierrez-Martinez et al, 2021 ). Thus, in the analytical review performed on the databases of Google Scholar, PubMed, IEEE Xplore and Elsevier Science Direct, it was shown that hDL-based BCI, apparently, will help overcome a significant drawback of the EEG signal classification ( Alzahab et al, 2021 ).…”

“…Most of the studies used a Convolutional Neural Network-Recurrent Neural Network (CNN-RNN) architecture, and half of the studies used a small number of layers to achieve a compromise between network complexity and computational efficiency ( Fujiwara and Ushiba, 2022 ). Further, the review also shows that it is necessary to use the neuroimaging method in the BCI paradigm, such as functional near-infrared spectroscopy (fNIRS), which becomes highly informative one ( Gutierrez-Martinez et al, 2021 ). Also it is important to apply the new combinations of architectures, such as RNN and Deep Belief Network based on DBN, for a better study of the frequency and time-frequency characteristics of the recorded brain signals.…”

Research hotspots and trends of brain-computer interface technology in stroke: a bibliometric study and visualization analysis

“…With advances in computer technology and neuroscience, brain-computer interface (BCI) technology has been receiving more and more attention and research in the treatment and rehabilitation of stroke patients ( Mane et al, 2020 ; Liu et al, 2022 ). BCI technology is an emerging technology that connects the human brain with external devices, using the neural activity of the human brain to control external devices for interaction and control, thereby giving patients a sense of autonomous control and accelerating the rehabilitation process ( Gutierrez-Martinez et al, 2021 ; Gao et al, 2022 ). Supported by BCI technology, nerve regeneration and functional recovery can be promoted, improving the quality of life and alleviating the impact caused by the disease ( Sinha et al, 2021 ; Song et al, 2022 ).…”