Video segment
Video of the patient’s myorhythmia while recumbent on an exam table and while seated in a wheelchair.
Figure 1
Effect of analysis parameters on time resolution, frequency resolution, and coherence estimates.
The 60 Hz sampling frequency of our IMUs provided a time resolution of 1/60 = 0.017 s. The first principal component of x-y-z acceleration and angular velocity recordings was interpolated to 1000 points per second using a MATLAB spline function to more accurately locate zero axis crossings (red +) without distorting the waveform (left column). The frequency resolution of a Welch Fourier spectrum is determined by the size of the segments in which the time series is divided. Smaller segments produce smoother (less noisy) spectra but with wider spectral peaks (middle column). Thus, the upper and lower limits of the half-power frequency band (red o) moved closer together when the segment size was increased from 512 to 1762, but the spectra computed with smaller segments were less noisy. Similarly, the statistical significance of coherence estimates (right column) depends strongly on the number of segments (i.e., length of segments). The 95% significance threshold (horizontal red line) moves lower when the time series is divided into a greater number of smaller segments, and the overall values of coherence and number of spurious peaks decrease. Color scheme of the power spectral plots in middle column: x – red, y – green, z – blue, and resultant – black.
Table 1
Tremor rhythmicity metrics for angular velocity recordings.
| RECORDING CONDITION | BODY PART | FREQUENCY (Hz) | MEAN PEAK-TO-PEAK AMPLITUDE (DEGREES) | TREMOR STABILITY INDEX (Hz) | HALF-POWER BANDWIDTH (Hz) | HALF-POWER BANDWIDTH INDEX (Hz) | HARMONIC DISTORTION (ANGULAR VELOCITY) | HARMONIC DISTORTION (ROTATION) |
|---|---|---|---|---|---|---|---|---|
| Rest | Head | 2.90 | 2.01 | 0.09 | 0.09 | 0.03 | 0.12 | 0.06 |
| R hand | 2.87 | 14.16 | 0.11 | 0.09 | 0.03 | 0.29 | 0.15 | |
| L hand | 3.05 | 6.51 | 0.15 | 0.15 | 0.05 | 0.17 | 0.09 | |
| Rest counting backwards | Head | 2.81 | 2.02 | 0.23 | 0.06 | 0.02 | 0.22 | 0.11 |
| R hand | 2.78 | 22.56 | 0.10 | 0.06 | 0.02 | 0.36 | 0.15 | |
| L hand | 3.19 | 13.37 | 0.38 | 0.09 | 0.03 | 0.26 | 0.12 | |
| Posture | Head | 2.58 | 13.27 | 0.15 | 0.21 | 0.08 | 0.11 | 0.06 |
| R hand | 2.52 | 60.63 | 0.13 | 0.09 | 0.04 | 0.32 | 0.14 | |
| L hand | 2.55 | 27.46 | 0.36 | 0.21 | 0.08 | 0.35 | 0.17 |
Figure 2
Gyroscopic recordings and power spectra for the head and right hand.
Gyroscopic recordings (first 4 seconds) and power spectra are shown for the head (left column) and right hand (right column). The patient was asked to relax while lying supine on an exam table. The power spectra contain a large peak at the fundamental frequency (2.89 Hz) and small harmonic peaks at 5.78 Hz and 8.67 Hz. The harmonic peaks in the head spectrum are smaller because the angular velocity waveform is more sinusoidal. The x-y-z axis data are shown in red, green and blue, and the resultant spectral power is shown in black.
Figure 3
Time-frequency power spectra and coherence plots of gyroscopic recordings.
Time-frequency waterfall plots of x-y-z spectral power are shown for the head (top row) and right hand (middle row). The patient was asked to relax while lying supine on an exam table. Time-frequency color maps of head vs right hand coherence are shown in the bottom row. Note the considerable variation in peak spectral power over time. Coherence of 0.8–1.0 occurred intermittently at the fundamental frequency (2.89 Hz) and less so at the harmonic frequencies.