Figure 1.

Experimental design and EEG processing. (A) BCI intervention protocol. (Top panels) BCI system design. Patients performed motor imagery tasks. Contralesional EEG signals were translated into commands to open or close the orthosis, which then provided proprioceptive sensory feedback to the patient as they performed motor imagery tasks. (Bottom panels) Intervention timeline. Patients were screened for the ability to perform the BCI task. Following screening, eligible patients underwent motor assessments before initiating the BCI therapy. Daily BCI therapy sessions included one calibration period (extended rest, alternating motor imagery and rest trials) and five BCI therapy runs (motor imagery and rest trials with active orthosis). Fidget periods were included between trials encouraging patients to blink or make physical adjustments. Motor assessments were performed every 4 weeks. Final EEG recording and motor assessment data were acquired after 12 weeks of therapy. (B) Data processing schematic for calculating PAC. The raw EEG signal was bandpass filtered in the lower (theta, alpha or beta) frequency range (right), and in the higher (high gamma) frequency range (left). Then, the complex analytic form of each signal was obtained using the Hilbert transform. The phase (angle of analytic signal) and power (amplitude of analytic signal) information was extracted from the lower- and higher-frequency signals, respectively. The coupling between phase and amplitude was then quantified using MVL algorithm to produce a modulation index values.