SKNA Signal Acquisition

1. Required hardware characteristics

   • Recording Protocol

SKNA is mostly captured from subclavian areas and left lower Abdomen, and also from the arm [4].

• Standard ECG during CPT [5]

• FastFix patch (fig. 5) placed vertically on sternum and SKNA recorded at 1-minute intervals during 3-minute standing and after sitting down for patients with chronic orthostatic intolerance [2]

b. Amplifier and Sensor Characteristics

• ML138 Bioamplifier [5]:
• Sampling Rate: 4/10 kHz
• Bandwidth: 5KHz
• Bittium Faros 180 (Bittium FastFix patch Electrode) [2]:
• Sampling Rate: 1 kHz (1000 Hz sampling rate allows the device to adequately sample signals with frequencies up to 500 Hz (Nyquist theorem))
• Resolution: 16 bit

2 extra Covidien patch electrodes connected by ME6000 biomonitor (sampled at 10,000 Hz)

• Portable NeuECG [1]:
  • Sampling frequency: 2/4/8 kHz
  • Ag/AgCl electrodes + conductive adhesive
  • Amplifier: ADS1299 module of Texas Instruments
  • ADC resolution: 24 bit
• STM32F103 microcontroller from STMicroelectronics for the MCU
• Number of channels: 8
• ADC Resolution: 24 bit
• Memory capacity: 16 Gb for TF card
• Battery capacity: 3000 mAh

Simultaneous recording by PowerLab Data Acquisition Hardware Device (ADInstruments) with the same SF, time, and BPF

• MP35 Biopac [6]
  • FS= 10kHz, duration: 10 min, participants: patients in ICU
  • Conventional ECG electrode

Hardware comparison: weight/monitoring time/portable/ noise level/ cloud/ real-time

2.2.  Physiological feature extraction

2. Signal processing

2.1. Signal preprocessing 

Because the greatest frequency components of the SKNA signal are about 1000 Hz, the “Nyquist frequency” rule dictates that a sampling frequency of around 2 kHz is required. [7] suggested a sampling rate of 10 KHz, but others, like [1], digitized signals at 2, 4, or 8 KHz.
Preprocessing stages are as below:

• 05-150 to display ECG
• Bandwidth: 0.05-1000 Hz via bandpass filter [1, 2]
• 50-70 Hz Bandstop (8-order Butterworth) filter to reject A/C interferences
• Bandpass filter (500-1000 Hz), (8-order Butterworth)
  • The ME6000 signals were bandpass filtered (500–1000 Hz) while the Faros 180 signals were high-pass filtered (300 Hz) to display SKNA.
• Summation of instantaneous SKNA over 100-ms time window = integrated SKNA (iSKNA) [6]
• Normalization [6]
• Signal rectification to nerve discharge assessment
• Moving Average (MA) to create a signal envelope
• Downsampled 200 times[1]
• RMS calculation as eSKNA [1]
• Signal Smoothing-Root Mean Square to show the strength of the signal

Binary time series graph and Intensity graph to highlight the activation time and lack of nerve activity

another method for determining threshold:

Two GUIs are used in order to process and visualize SKNA trend, including Visual programming environment: Labview, NerveAct [5] and LabChart 8 Pro software [2] LabChart pro 8 software (ADInstruments) [1].
Validation of SKNA is almost always done by BPM (inverse HRV) [5], in addition for verification data were recorded by Powerlab as a standard system simultaneously [1].SKNA is mostly evaluated as onsets and offsets concurrent with start and stop time of the experiment which can be coincident with BPM acceleration as well. Correlation, baseline (lower means better noise cancelation level) and frequency spectrum (Powerlab as a standard system has higher energy in 1000-2000 Hz which means higher noise level) are among evaluation parameters[1]. Also, one study has investigated the structural complexity changes of iSKNA by calculating Fractal structure.
For analyze these structures Multiscale fluctuation Analysis (MSFA) is used which iteratively calculates the standard deviations of the residuals after removing the linear-fitted trend of each scale from the normalized iSKNA [6].
The α exponent, the slope of the curve in log-log plot of fluctuations of iSKNA, indicates fractal structure of SKNA. They used intrinsic fluctuation of SKNA as an indicator to mortality rate estimation for patients in critical care unit and showed the higher intrinsic fluctuation of SKNA at 60 s (the responsiveness of the sympathetic nerve activity to environmental challenges) can be an independent predictor for both 30- and 90-day mortality [6].
The other features are the average voltage of SKNA (aSKNA), a variable value of SKNA (vSKNA), and the number of SKNA bursts (bSKNA) are used to evaluate SKNA[8].
Integrated SKNA was calculated as the sum of instantaneous SKNA across a 100-ms time window (iSKNA). The recording’s averaged iSKNA was employed as a metric to indicate total sympathetic nerve activity[7].
The iSKNA fluctuations at different time scales are normalized to avoid the effects of overshooting neuronal activity, noise, and between-subject changes in iSKNA amplitude.

2.3. The Result of Our Test
The result of a real CPT (Cold Pressure Test) test can be seen in the following.
The recorded ECG data and its high frequency components is shown in figure 9.

Signal processing:

After powerline noise canceling and applying a bandpass filter (0.5-50 Hz) one can see the ECG signal.

Heart Rate Variability (HRV) and Beat Per Minute (BPM) is calculated using ECG. SR refers to Sampling Rate which is 2kHz.

BPM = 60 * SR / HRV

SKNA Acquisition:

1. Rectification
2. Windowed Mean Calculation (mSKNA)
3. Signal envelope Calculation (eSKNA)

The raw signal and the output of each step is illustrated successively:

Conclusion:

By comparing eSKNA and BPM signals, it can be concluded that the sympathetic activity is enhanced during CPT and consequently led to BPM increasement.

References

[1]        Y. Xing et al., “A Portable NeuECG Monitoring System for Cardiac Sympathetic Nerve Activity Assessment,” in 2020 International Conference on Sensing, Measurement & Data Analytics in the era of Artificial Intelligence (ICSMD), 2020, pp. 407-412: IEEE.

[2]        X. Liu et al., “Using an ambulatory electrocardiogram monitor to record skin sympathetic nerve activity,” Heart Rhythm, 2021.

[3]        C. Liu and J. Li, Feature Engineering and Computational Intelligence in ECG Monitoring. Springer, 2020.

[4]        T. H. Everett IV, A. Doytchinova, Y.-M. Cha, and P.-S. Chen, “Recording sympathetic nerve activity from the skin,” Trends in cardiovascular medicine, vol. 27, no. 7, pp. 463-472, 2017.

[5]        C. Liu, J. Wong, A. Doytchinova, P.-S. Chen, and S.-F. Lin, “Method for detection and quantification of non-invasive skin sympathetic nerve activity,” in 2018 International Conference on System Science and Engineering (ICSSE), 2018, pp. 1-4: IEEE.

[6]        J.-J. Chen et al., “Complex dynamics of skin sympathetic nerve activities as a prognostic predictor for critically ill patients,” Journal of the Formosan Medical Association, vol. 120, no. 1, pp. 660-667, 2021.

[7]        J.-J. Chen et al., “Complex dynamics of skin sympathetic nerve activities as a prognostic predictor for critically ill patients,” Journal of the Formosan Medical Association, vol. 120, no. 1, Part 3, pp. 660-667, 2021/01/01/ 2021.

[8]        P. Zhang et al., “Characterization of skin sympathetic nerve activity in patients with cardiomyopathy and ventricular arrhythmia,” Heart rhythm, vol. 16, no. 11, pp. 1669-1675, 2019.