Fig. 4: Sonomyographic pattern of cough.

A Example B-mode ultrasound and SMGs from one participant during a cough. The images and SMGs were obtained synchronously from multi-view system. The quintic Savitzky–Golay filter with a window length of eleven was used to smooth the signals for better illustration. The white markers inside the ultrasound images indicate the distance representation of the target layers or the direction of displacement. B Correlation analysis of thickening fraction (\({\mathrm{TF}}\), Corr = 0.69, p < 0.001). C Correlation analysis of peak thickening speed (\({{\mathrm{TS}}}_{{\mathrm{peak}}}\), Corr = 0.67, p < 0.001). D Correlation analysis of average thickening speed (\({{\mathrm{TS}}}_{{\mathrm{ave}}}\), Corr = 0.50, p < 0.001). E Correlation analysis of thickness at end-compression (\({T}_{{\mathrm{com}}}\), Corr = 0.50, p < 0.001). F Correlation analysis of relative thickening speed (\({{\mathrm{TS}}}_{{\mathrm{rel}}}\), Corr = 0.48, p < 0.001). G Correlation analysis of passive excursion (\({\mathrm{PExc}}\), Corr = 0.36, p < 0.001). H Correlation analysis of peak velocity (\({V}_{{\mathrm{peak}}}\), Corr = 0.32, p < 0.010). I Correlation analysis of average velocity (\({V}_{{\mathrm{ave}}}\), Corr = 0.31, p < 0.010). Eight SMG features that have significant correlations with cough peak flow (p < 0.050) are shown in the figure and ordered by correlation coefficient. Features from some unmeasurable samples were dropout due to unsatisfactory ultrasound imaging quality. Corr: Pearson correlation coefficients. For the scatter plots in B–I, the distribution graphs in the top and right margins correspond to the parameters on the horizontal and vertical axes, respectively. The center lines and shaded regions represent regression lines calculated by the least squared method, along with its 95% confidence intervals. The two-sided Pearson correlation analysis was used across all participants (n = 60) to assess if there is a significant correlation.