Exploring motor imagery–induced neural activation and corticospinal excitability in healthy adults using EEG and TMS

Abstract

Motor imagery (MI)–based brain–computer interfaces (BCIs) activate motor–related brain regions and, through neurofeedback, foster neuroplasticity, offering significant potential for neurorehabilitation [1]. While MI–BCIs have shown success in restoring hand function after stroke, their use in mitigating upper–limb impairments in multiple sclerosis (MS)—a chronic neurodegenerative disorder characterized by impaired motor control and coordination—remains underexplored [2]. This study presents a preliminary investigation into the feasibility of using MI–BCI for targeted therapy of two specific hand motor function deficits common in MS: fine motor control, represented by a task involving repeated cycles of full hand extension and thumb–fingertip closure, and motor coordination, represented by a sequential finger–to–thumb opposition task. Using two hand tasks representing each of these functions, the objectives were: 1) to examine whether MI of these tasks alters corticospinal excitability (CSE), indicating the potential for fostering neuroplasticity and improving hand function, and 2) to determine if distinct task–specific neural activation patterns can be detected via electroencephalography (EEG), suggesting they may be targeted with BCI–based neurofeedback training. Data from twenty–one healthy participants (8 males, 13 females; mean age: 41.35 ± 8.36 years) were analyzed for this study. Transcranial magnetic stimulation (TMS) was used to assess changes in CSE due to MI of the coordination and control tasks. A single TMS pulse was delivered during intervals of MI and rest and resulting motor–evoked potentials (MEPs) were measured from the first dorsal interosseous (FDI) muscle. For the control task, MI did not significantly increase MEP amplitudes compared to rest (Δ=39.91 μV, p=.354), but significantly decreased MEP latencies (Δ=−0.34 ms, p=.002). For the coordination task, MI also did not significantly increase MEP amplitudes (Δ= 29.67 μV, p = 0.243), but significantly decreased MEP latencies compared to rest (Δ=−0.30 ms, p = 0.006). CSE was also assessed during actual execution of the tasks (ME), and changes in both MEP amplitudes and latencies were significantly different from rest for both tasks (control: Δ=515.96 μV, p < .001; Δ=−1.27 ms, p < .001; coordination: Δ=534.98 μV, p < .001; Δ=−1.15 ms, p < .001). In a separate session, 64–channel EEG was recorded as participants performed 60 intervals each of MI and ME of the two tasks, as well as rest. To determine if the control and coordination tasks were distinguishable using EEG, several different classification pipelines were explored. In terms of feature extraction, Filter Bank Common Spatial Patterns (FBCSP) based on the five standard EEG frequency bands (delta, theta, alpha, beta, gamma), as well as on a finer set of nine overlapping bands spanning 1–40 Hz in 4 Hz increments, was explored, as were functional connectivity–based features, specifically correlation, coherence, and phase–locking value (PLV). Classification was performed using multiple algorithms, including Linear Discriminant Analysis (LDA), Support Vector Machines (SVM), and Decision Trees (DT). Using the PLV features, classification of control vs. coordination tasks yielded high accuracy (80.5% for MI and 82.7% for ME using a SVM classifier). Accuracies above the chance threshold (i.e., 58.3% for n=120 trials, α=0.05, based on binomial distribution) were obtained for all subjects for both MI and ME, and a majority of participants achieved accuracies above 70%—19 participants for MI and 17 for ME— indicating individual potential for task–specific decoding. Although MI did not significantly increase MEP amplitude, the consistent latency reductions suggest enhanced corticospinal conduction, likely reflecting subthreshold motor activation. These results may support the potential of MI to facilitate neuroplasticity, despite limited task–specific modulation in excitability. On average, the distinct neural patterns differentiating motor control and coordination were detected with accuracy greater than chance, and for some participants with very high accuracy. Together the EEG and TMS results support the feasibility of targeted BCI–based therapy for improving motor control and coordination. Further work is needed to more reliably and specifically identify the neural activation patterns associated with these functions, particularly in people with MS, and then to evaluate the efficacy of BCI–mediated MI therapy for improving upper limb function in the target population.