Mu-Beta event-related (de)synchronization and EEG classification of left-right foot dorsiflexion kinaesthetic motor imagery for BCI

PLOS ONE RESEARCH ARTICLE Mu-Beta event-related (de)synchronization and EEG classification of left-right foot dorsiflexion kinaesthetic motor imagery for BCI Madiha Tariq, Pavel M. Trivailo ID, Milan Simic ID* School of Engineering, RMIT University, Melbourne, VIC, Australia * a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Citation: Tariq M, Trivailo PM, Simic M (2020) MuBeta event-related (de)synchronization and EEG classification of left-right foot dorsiflexion kinaesthetic motor imagery for BCI. PLoS ONE 15 (3): e0230184. https://doi.org/10.1371/journal. pone.0230184 Editor: Dhakshin Ramanathan, University of California San Diego, UNITED STATES Received: June 13, 2019 Accepted: February 24, 2020 Published: March 17, 2020 Copyright: © 2020 Tariq et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Abstract The left and right foot representation area is located within the interhemispheric fissure of the sensorimotor cortex and share spatial proximity. This makes it difficult to visualize the cortical lateralization of event-related (de)synchronization (ERD/ERS) during left and right foot motor imageries. The aim of this study is to investigate the possibility of using ERD/ERS in the mu, low beta, and high beta bandwidth, during left and right foot dorsiflexion kinaesthetic motor imageries (KMI), as unilateral control commands for a braincomputer interface (BCI). EEG was recorded from nine healthy participants during cuebased left-right foot dorsiflexion KMI tasks. The features were analysed for common average and bipolar references. With each reference, mu and beta band-power features were analysed using time–frequency (TF) maps, scalp topographies, and average time course for ERD/ERS. The cortical lateralization of ERD/ERS, during left and right foot KMI, was confirmed. Statistically significant features were classified using LDA, SVM, and KNN model, and evaluated using the area under ROC curves. An increase in high beta power following the end of KMI for both tasks was recorded, from right and left hemispheres, respectively, at the vertex. The single trial analysis and classification models resulted in high discrimination accuracies, i.e. maximum 83.4% for beta ERS, 79.1% for beta ERD, and 74.0% for mu ERD. With each model the features performed above the statistical chance level of 2-class discrimination for a BCI. Our findings indicate these features can evoke left-right differences in single EEG trials. This suggests that any BCI employing unilateral foot KMI can attain classification accuracy suitable for practical implementation. Given results stipulate the novel utilization of mu and beta as independent control features for discrimination of bilateral foot KMI in a BCI. Data Availability Statement: All relevant data underlying the study is available within the manuscript and its Supporting Information files. Funding: The author(s) received no specific funding for this work. 1. Introduction Competing interests: The authors have declared that no competing interests exist. People affected by neurological disorder, stroke, or spinal cord injury (SCI) necessitate a therapeutic goal of motor gait rehabilitation using assistive technologies [1, 2]. For lower-limb PLOS ONE | https://doi.org/10.1371/journal.pone.0230184 March 17, 2020 1 / 20 PLOS ONE Mu-Beta event-related (de)synchronization and EEG classification affectees, to re-gain the dorsiflexion of foot drop is vital [3–5]. The lost motor control functions could be emulated by inducing neuroplasticity using a brain-computer interface (BCI) system [6]. BCI provides an alternative neuropathway that translates human brain activities into commands for controlling external devices or prostheses [6, 7]. BCIs that use EEG features such as oscillatory/sensorimotor rhythm (SMR) are recorded over the somatic sensorimotor cortex. SMR are concentrated in the alpha (mu) (7–12 Hz), beta (13–35 Hz), and gamma (>36 Hz) frequency bands [8, 9]. BCIs have successfully deployed SMR to identify any changes related to the physical movement (motor execution, ME) or imagination of movement (motor imagery, MI) of any limb [10]. This is because an increase in the corticomotor excitability is involved during MI and ME of limb movement which is both muscle-specific and temporally modulated [11]. Both the execution and imagery tasks have been used in experiments, because the ME and MI implicate overlapping neural structures within the central nervous system [11]. However, from literature, MI tasks have been preferred over ME ones, to avoid any possibility of proprioceptive feedback. The MI is a covert cognitive process, where the user makes a kinaesthetic imagination of his/her own limb movement without any muscular intervention, also called kinaesthetic motor imagery (KMI) [1, 12]. Each limb movement elicits a unique pattern in the SMR mu and beta features [9]. These patterns are reflected in the form of either a power decrease termed event-related desynchronization (ERD) that correlate to movement preparation [13], or an increase in power termed event-related synchronization (ERS) associated to resting/idling state, or an inhibitory state [14, 15]. Increased idling might correspond to reduced inhibition of downstream, movement related neurons. The cortical localization of ERD/ERS patterns is due to the somatotopic arrangement of the motor cortex. The upper limbs e.g. hand area representation is on the mantle of the cortex, followed by lateralization [16], that makes the spatial discrimination between left and right movement prominent compared to lower limbs. The right-left hand ME or KMI mu ERD correlate to the bilateral hand area (C3 and C4 electrode positions) of the sensorimotor cortex with evident contralateral dominance compared to ipsilateral side [17, 18]. These contralateral and ipsilateral differences in mu ERD have been classified by BCI to be used as control features for operating external devices [19–21]. On the contrary, right-left lower-limb ME or KMI tasks are not extensively deployed due to the close location of left-right lower-limbs’ areas to each other. The foot motor area is situated deep within the interhemispheric fissure of the sensorimotor cortex that makes the left and right foot ME or KMI difficult to be spatially discriminated since they produce nearly identical EEG patterns [16]. Therefore, we can find studies where general foot KMIbased BCIs deploy feet KMI as one feature without discriminating the left-right side [22, 23]. However, studies available on the left-right discrimination of foot KMI, proposed the mu ERD and beta ERS/rebound (post task completion), as possible EEG features for classification [7, 17, 24, 25]; where the ERD/ERS patterns generate at the vertex [24]. According (...truncated)