Fig. 4
From: An objective method to detect sighs during cardio-pulmonary exercise testing
The same patient is represented as in Fig. 3B. In (A), the purple line represents the mean filter of 15 tidal volume values. In figure (B), the orange line represents the median filter of 7 values filter. In both A and B, the green line represents the median filter of 15 values. Each grey dot represents one respiratory cycle. In figure (A) and (B), detected spike of VT are not shown because two filters are represented. Figure (A) illustrates that the mean filter (purple line) is more influenced by outliers of VT than the median filter (green line) for a same window size (15). The detection of sighs using the mean filter of 15 values is of only 4 spikes of VT detected in this patient. This contrasts with Fig. 3B from the manuscript using a median filter of 15 values and detecting 17 spikes of VT. In figure (B), the frequency of the sighs is so high (more than 1 sigh every 2 breaths on a sequence of 7 breaths) that the centred value of the 7 points median filter reaches the value of the sighs two times as shown by the two spikes of the orange line. When the value of the filter represents the value of the sigh, the method will be unable to detect sighs. The median filter of 7 values is globally less stable than the median filter of 15 values which globally follows the underlying breathing without being influenced by the sighs. While this extreme situation of instability is infrequent, it leads to less sighs’ detection using a median filter using smaller window of filtering (in this example 7 compared to 15).
