Wavelet Analysis of Respiratory Muscle sEMG Signals during the Physiological Breakpoint of Static Dry End-Expiratory Breath-Holding in Naive Apneists: A Pilot Study

Figures

• 

  Figure 2. Placement of Delsys Trigno wireless sensors for unilateral right-sided body data collection of the myoelectric activity for the three respiratory muscles—the accessory inspiratory muscle scalenus anterior at medium (SC; green), the primary inspiratory muscle parasternal intercostal (IC; magenta), the accessory expiratory muscle rectus abdominis (RA; cyan), and for the locomotor muscle brachioradialis (BR; black). The colors of the sensors correspond to the colors of the time markers (T1start/T1stop and T2start/T2stop, Figure 1) on all signal plots in the time domain.

• 

  Figure 4. Time series of sEMG signals for three respiratory muscles EMGSC, EMGIC, and EMGRA and one nonrespiratory muscle EMGBR, respectively, from top to bottom, which are given for a representative apneist BHr3 (BHD Rank #1; Table 1). The first and second BH phases (BH1 and BH2) are marked with time markers (T1start/T1stop and T2start/T2stop, Figure 1) and colored rectangles, while their colors correspond to the sensor colors in Figure 2. In the final stages of BH, a gradual increase in the magnitudes of EMGSC, and especially EMGIC, was noticeable in both BH1 and BH2 periods.

• 

  Figure 5. Time series of EMGSC, EMGIC, and EMGRA, and EMGBR, respectively, from top to bottom, which are given for a representative apneist BHr4 (BHD Rank #1, highest BHD score; Table 1). Other descriptive notes in the caption for Figure 4. In contrast to BHr3, limited or no changes were observed in EMGSC and EMGIC; however, abrupt changes in EMGRA magnitudes were present in the final stages of BH2, indicating a different BH response mechanism.

• 

  Figure 6. Comparison of the occupied bandwidth (OBW) estimates (the width of the frequency band that contains 99% of the power of the signal, the lower and upper bounds of the band, and the power in the band), of the muscle sEMG signals in the initial and final phases of breathholding (during BH1 and BH2). The lower bounds were about 7 Hz, which were a consequence of an AC-coupling filtering with 7 Hz highpass filter that strongly attenuated lower frequencies in the sEMG signals. As a result, the upper bounds were similar to the occupied bandwidth. The increase in occupied bandwidth was most pronounced for IC, followed by SC. The occupied band power was the highest for RA and the lowest for BR, which indicated that it was necessary to normalize the power in order to be able to compare the change in the initial and final BH phases for different muscles.

• 

  Figure 9. Correlations of neuromuscular coupling for inspiratory muscles SC and IC were demonstrated by comparing scalograms for EMGSC and EMGIC signals of naive apneist BHr12 (BHD Rank #3). Correlation of areas of high energy density for EMGSC and EMGIC signals in the T-F plane were characteristic for the approximate FB of 120–480 Hz. This was the same FB for the increased energy density of the BH final stages for EMGIC signals in both BHr12 and BHr3, as shown in Figure 7, thus indicating that this FB is characteristic of increased IC electromyographic activity, regardless of the primary stimulus.

• 

  Figure 10. Individual specificities of neuromuscular coupling in participants for SC were demonstrated by comparing the scalograms for EMGSC signals of apneist BHr6 (BHD Rank #1) with BHr4 (BHD Rank #1, which are presented in Figure 8). During the BH maneuver, participant BHr4 (rower) showed higher SC activity during spontaneous breathing (areas of high energy density for EMGSC signal in the T-F plane are related to inspiration), while BHr6 (diver) showed greater SC activity during breath holding.

• 

  Figure 11. Joint distribution of relative energies of wavelet MRA (WMRA) components of respiratory signals EMGSC, EMGIC, and EMGRA of all participants for three types of data grouping: professional/amateur (left), respiratory signal type (middle), and MRA components (right). The distributions were determined using a probability density estimate based on a normal kernel function. The lines of best fit of the two data sets, professionals and amateurs, are calculated using linear regression, where r is the appropriate estimate of the regression slope.

• 

  Figure 12. Separated distributions of relative energies of WMRA components of respiratory signals EMGSC, EMGIC, and EMGRA of all participants for professional/amateur grouping. As in the caption of Figure 11, the lines of best fit of the two data sets, professionals and amateurs, are calculated using linear regression, where r is the appropriate estimate of the regression slope. A comparison of the distribution curves for professionals and amateurs shows that they most closely matched for the IC.

• 

  Figure 14. Boxplot representation of the changes in relative energy across the scales △P(ε) between the BH final and initial stages (as explained in the caption of Figure 13), calculated for the combined BH1 and BH2 periods. The changes were most pronounced for the IC muscle, in particular an increase of D2 and D3 components, with a synchronous decrease of the D6 component (compare with the captions of Figure 13 and Figure 15).

• 

  Figure 15. Changes in relative energy across the scales △P(ε) between the BH final and initial stages calculated for both BH1 and BH2 periods (left) and corresponding to the average values per scale for the total population and the population grouped by ranks (right). Scales with a significant positive change are displayed in boldface, and with negative change are italic (all values are expressed in percentages). Gradual highlighting from negative (blue) to positive (red) values was performed at the level of one muscle. In the part of the figure where the average values are calculated, the width of the red and blue boxes visualizes the size of the positive and negative change in values, respectively. Average values show the largest energy increase for D2 and D3 scales of EMGIC and EMGSC signals, regardless of the rank, as well as for D3, D4, and D5 scales of EMGRA signal, but only for the Rank #1 group.

• 

  Figure 1. Breathing pattern of breath-holding (BH) maneuver. BH began after completion of spontaneous exhalation to functional residual capacity and ended with spontaneous inhalation to normal lung capacity. The T1start/T1stop and T2start/T2stop markers define the beginning/ending of the first and second exhalation BH, respectively (the time markers are displayed on all signal plots, whether in the time or wavelet domain, while their colors correspond to the sensor colors in Figure 2).

• 

  Figure 3. The equivalence of processing for a discrete time signal x={xn}n=1N between DWT using wavelet and scaling functions ({Ψj}j=15 and Φ5) and the corresponding fDWT using lowpass and highpass filters (L and H). Left: A perfect T-F plane tiling for DWT in dyadic (octave range) configuration, thus resulting in Heisenberg boxes with the same constant area (examples highlighted in red and blue). With successive downscaling, the wavelet is scaled by 2−1 and translated by 2. Right: The corresponding fDWT bandwidth division using the cascade in the form of a tree structure of halfband filters and subsampling by 2 results in filterbank with a constant relative bandwidth. Such filtering followed by sampling at the respective Nyquist frequency fs/2 gives same wavelet and smooth coefficients ({Dj}j=15 and S5) as for DWT. More detailed explanations in the main text.