Second heart sound splitting as an indicator of interventricular mechanical dyssynchrony using a novel splitting detection algorithm

Figures

• 

  Figure 1. Demonstration of S‐transform for signal analysis using an S2 from an animal experiment. Panel A shows (from bottom to top) 1) the S2 during biventricular pacing with the right atrium to RV (A‐RV) paced interval 150 ms and A‐LV 50 ms, 2) a mother wavelet of 100 Hz and the resulting amplitude values with time, and 3) a mother wavelet of 200 Hz and the resulting amplitude values with time. Panel B shows the complete S‐transform amplitude spectrum. In panel A, the mother wavelets of various frequencies slide along the heart sound signal, convolve with the signal, and result in amplitude values. In panel B, amplitude values of frequencies from 50 Hz to 250 Hz are plotted as an intensity map

• 

  Figure 2. START algorithm for estimating A2‐P2 splitting interval. (a) A segmented S2 from animal experiment No.1 with A‐RV delay 150 ms and A‐LV delay 50 ms. (b) S‐transform amplitude spectrum and five ridges (solid yellow lines). The red number indicates the rank of each ridge. The white number indicates the weight factor of each ridge. The raw heart sound signal in panel A was processed with S‐transform to obtain a time–frequency–amplitude map in panel B, then the ridges of this map were identified and ranked according to their energies. The two ridges with the highest energies were used for calculating heart sound splitting interval. In this example, the START‐estimated S2 splitting interval is 44 ms

• 

  Figure 3. Demonstration of interventricular relaxation dyssynchrony calculation. (a) Raw LV and RV pressures from animal experiment No.1 at A‐RV delay 150 ms and A‐LV delay 50 ms. (b) Normalized LV and RV pressure, with their downslopes used for correlation calculation. (c) Correlation coefficient curve indicating an interventricular relaxation dyssynchrony of 11 ms. Raw LV and RV pressures in the panel A were normalized in the panel B. The downslopes of the normalized pressures were cross‐correlated to obtain the correlation coefficients in the panel C. The time when the correlation coefficient was the highest was used as an estimate of interventricular relaxation dyssynchrony. LVP =left ventricular pressure; RVP =right ventricular pressure

• 

  Figure 4. START algorithm in simulated second heart sounds. Red solid line indicates line of identity in panel A and 30 ms in panels B and C. Red dash lines are 5‐ms upper and lower boundaries of the expected values. Panel A (in blue) demonstrated three representative simulated heart sounds of splitting intervals 10 ms, 40 ms, and 70 ms, whereas the bottom figure showed the relationship between simulated and START‐estimated splitting intervals from 10 ms to 70 ms. Pearson's correlation was calculated. Panel B (in green) demonstrated three representative simulated heart sounds with normalized A2 amplitudes of 0.2, 2.0, and 4.0 when the normalized P2 amplitude was fixed at 1.0. In all these cases, the simulated splitting intervals were fixed at 30 ms, and the bottom figure showed the errors of splitting interval estimation using our proposed START algorithm. Panel C (in magenta) demonstrated three representative simulated heart sounds of signal‐to‐noise ratios of 10, 20, and 30 dB. In all these cases, the simulated splitting intervals were fixed at 30 ms, and the bottom figure showed the errors of START‐based splitting interval estimation for signal‐to‐noise ratios from 10 dB to 30 dB at…

• 

  Figure 5. Representative examples of second heart sound splitting at various paced interventricular delays. When A‐RV delay was fixed at 150 ms, A‐LV delays were varied from 50 ms to 250 ms with 25 ms per step. Representative heart sounds from experiment No.1. In the left panels, three situations with A‐LV delays of 75, 150, and 225 ms are shown with recordings of electrocardiogram, LV pressure (blue), RV pressure (red), and heart sounds. In the right panel, S2 is shown at higher temporal resolution. Note that S2 splitting decreases from A‐LV delays of 50 ms to 150 ms, but remains virtually constant at longer A‐LV delays. S1 = the first heart sound; S2 = the second heart sound; LV =left ventricle; RV =right ventricle; A‐LV =right atrium to left ventricle paced delay.

• 

  Figure 6. Second heart sound splitting interval as a function of atrio‐left ventricular stimulation interval (A‐LV). Each color represents a different animal. Median values and 25–75% percentiles of all beats at each A‐LV interval are presented (A‐RV delay was fixed at 150 ms). From experiments No. 1 to 5, the median number of heartbeats was 38, 43, 34, 29, and 33, respectively, for each A‐LV delay group. Spearman's rank correlation R2 value was 0.71, 0.41, 0.63, 0.64, and 0.72 for experiments 1 to 5, respectively (p < .001 in all experiments). S2 = second heart sound

• 

  Figure 7. (a) Changes in interventricular relaxation dyssynchrony (IRD) as a function of A‐LV delay (n = 5). A‐RV delay was fixed at 150 ms. From A‐LV delay of 50 ms to 250 ms at step of 50 ms, the overall IRD was 8 (4,12) ms, 2 (−2,8) ms, −14 (−24, −5) ms, −20 (−39, −11) ms, and −18 (−28,‐11) ms, respectively. Data are presented as median (25 and 75 percentile). Spearman's rank correlation R2 value was 0.72, 0.73, 0.75, 0.64, and 0.82 for experiments 1 to 5, respectively (p < .001 for each experiment). (b) Relation between IRD and S2 splitting. Each point represents a heartbeat. Each color represents an experiment, the same as in Figure 6. Equations and Pearson's correlation coefficients of the correlation plots are summarized in Table 2). A‐LV =right atrium to left ventricle paced delay; S2 = the second heart sound