Impact of the auscultation area on heart sound waveforms for the estimation of Cardiac Time Intervals

Although clinicians still regard traditional auscultation as a key bedside assessment tool, interest in its digital counterpart has recently grown. Digitalizing heart sounds enables the extraction without the need of echocardiography of novel clinical indicators like Cardiac Time Intervals (CTIs), temporal parameters reflecting the hemodynamic behavior of the heart. The estimate of the CTIs grounds on the identification of the time of closure of heart valves, which the human ear cannot perceive - but algorithms can. Traditional auscultation areas are still considered a gold standard, but their suitability for CTI estimation was never thoroughly explored. In this study, we analyze the spatial variability of heart sound waveforms over the chest with high spatial resolution. We estimate the time of closure of cardiac valves and define a goal-oriented objective function to automatically identify the best auscultation area for each valve. We compare the selection against visual inspection (manually selecting the signals where discriminating between the left- and right-heart contributions was possible) and against the traditional auscultation areas. We found that the spatial similarity pattern of the sound waveforms recorded using the multi-source sensor is consistent with expectations. The variation in morphology produces a variation in the estimates, whose standard deviation ranges from 7 to 13 milliseconds, potentially causing issues in clinical interpretation. This proves that the selection of the recording point impacts the estimate and must be carefully defined. The estimates resulting from the automatic channel selection present higher correlation coefficients against visual selection (R = 0,73 to 0,96) than against traditional auscultation (R = 0,49 to 0,92). We can conclude that the definition of the best auscultation areas should be considered a goal-dependent task and that the spatial variability of heart sounds plays a role in obtaining robust estimates of clinical biomarkers.Clinical Relevance-The reported spatial variation of the heart sound waveforms makes the definition of the best auscultation areas a goal-dependent task that should be adapted to the clinical task.