Impact of Movement Amplitude in Dynamic Visual Cues on Motor Imagery Ability and BCI Classification
Description
Motor imagery-based brain-computer interface (MI-BCI) is an important research topic in the field of motor function rehabilitation. Dynamic visual cues can modulate motor imagery (MI) performance, yet the influence of movement amplitude remains underexplored. This study investigates the modulatory effects of movement amplitude in dynamic visual cues on motor imagery ability and MI-BCI classification performance. Participants performed right-hand MI tasks cued by dynamic visual stimuli of three amplitudes (large: 60°, medium: 40°, small: 20°). Subjective kinesthetic imagery vividness, beta-band Event-related desynchronization (ERD), and classification accuracy were measured. Results showed that large-amplitude visual cues significantly enhanced subjective vividness scores (KMI: 3.14 ± 0.85), elicited stronger MI-related cortical activation (beta-band ERD: -6.92 ± 9.97), and achieved higher MI-BCI classification accuracy (e.g., beta-band accuracy: 93.45% ± 5.37%) compared with medium and small amplitudes. These findings indicate that the amplitude of dynamic visual cues modulates both subjective and objective MI performance, providing practical insights for optimizing MI-BCI training paradigms and HCI-based rehabilitation interventions. During formal testing, the participants maintained physical stillness. Adopting the Graz training paradigm, each condition comprised 5 practice trials followed by 30 experimental trials (Yang et al., 2024). Before each MI block, participants practiced the tapping motion at the specified amplitude (20°, 40°, or 60°) guided by the experimenter. The experimenter demonstrated the correct angle using a goniometer and visually monitored the participant’s movements during practice to ensure compliance. A single trial lasted 13 seconds with four phases: Fixation (2 s): A black "+" at the center of the screen prompted attentional focus. Visual cue (1 s): A dynamic visual cue was presented on the screen. MI period (5 s): Participants repetitively imagined the right-hand palm-tapping sensation. Rest (5 s): A "relax" prompt appeared for recovery. After each block, participants completed the subjective KVIQ scale. Blocks lasted approximately 10 minutes with 2-minute inter-block rest periods. The ErgoLAB 3.0 software was used for paradigm presentation, EEG calibration and adjustment, and data acquisition. EEG signals were recorded using a 32-channel Bitbrain semi-dry electrode system (sampling rate: 256 Hz).
Files
Steps to reproduce
Thirty-two participants aged 19–25 years were initially recruited for this study, including 16 females (M = 22.70, SD = 1.26) and 16 males (M = 23.51, SD = 1.23). An a priori power analysis was conducted using G*Power for a repeated-measures ANOVA with three levels. With a medium effect size of f = 0.25, an α level of 0.05, and a desired statistical power of 0.80, the minimum required sample size was estimated to be 28 participants. Therefore, 32 participants were recruited to account for potential data loss due to EEG artifacts. All participants had normal sensory and motor function, were right-handed, and had no prior experience with similar experiments. Before each experimental session, participants were required to obtain at least 8 hours of sleep the previous night and to refrain from alcohol, caffeine, or tobacco consumption for 3 hours prior to the experiment. Prior to the formal experiment, participants completed a basic demographic survey and received detailed instructions regarding the experimental procedures and measurement protocols. All participants voluntarily signed written informed consent forms. EEG data were collected from all 32 participants. Four participants were excluded because of excessive motion artifacts that compromised EEG signal quality, resulting in 28 valid datasets for the final statistical analyses. Thus, the final valid sample size still met the minimum sample size estimated by the a priori power analysis. The experiment followed a within-subjects design, with all participants exposed to three dynamic visual-cue conditions. The three conditions were counterbalanced using a Latin square order. The total task duration was ≤ 60 minutes, and the experimental procedure is illustrated in Figure 1. The experiment was conducted in a controlled-noise environment (doors/windows closed) at Tianjin University’s Industrial Design Laboratory. Upon arrival, participants received instructions about the protocol, provided demographic information, and signed informed consent. The cardboard box used in the recording of the stimulus materials was placed to the right of the computer as part of the experiment. Before MI tasks, participants sat in a comfortable chair 60 cm away from the screen, with their right palm naturally placed on the cardboard box. Participants were explicitly instructed to adopt a KMI strategy: they were asked to “feel the sensation of your right hand tapping the box, as if you were actually performing the movement, rather than visualizing yourself doing it from an external perspective.” To ensure compliance, after each block, participants were asked to briefly describe their imagery strategy. An EEG cap was fitted and calibrated prior to formal testing.
Institutions
- Tianjin UniversityTianjin, Tianjin