Decoding of Human Movements Based on Deep Brain Local Field Potentials Using Ensemble Neural Networks

Figure 2

(a) Instantaneous amplitude (left) and spectrogram (right) of the right-sided STNs LFP for all frequency bands of patient #1 in a 4-second window centered at the time of response and visually cued left finger clicking events; the extracted frequency bands of LFP signal are delta (0–4 Hz), θ (4–8 Hz), α (8–12 Hz), low β (13–20 Hz), high β (20–32 Hz), low γ (32–60 HZ), and high γ (60–90 Hz), respectively, where high γ band [11] is not the same as the conventional high gamma band (80–200 Hz). (b) The instantaneous magnitude of different bands computed using Hilbert Transform (HT) for all trials of patient #1 during left and right finger visual cued clicking events obtained from deep brain’s left STN LFPs (motor responses situated at the center of each time scale) [12]. (c) The average instantaneous magnitude (blue line) and standard deviation (SD) (green shadow area) acquired from STN LFPs of each component for patient #1 and visual cued left and right finger clicking events within 2 s time window. For each frequency band, LFP signal features were defined with average amplitude in five segments (area covered by dotted line.)

(a)

(b)

(c)