Variance fractal dimension trajectory as a tool for heart sound localization in lung sounds recordings

Methods by which to deduce information regarding respiratory mechanics via respiratory acoustics would present favorable additions to traditional pulmonological testing for determination of respiratory conditions. The combination of sounds in sound signals acquired on the chest wall, mainly those originating from pulmonary airflow and the heart, complicate the establishment of a firm definition of flow-specific lung sounds as a function of airway narrowing. In this study, the variance fractal dimension trajectory (VFDT) algorithm has been applied to lung sounds data to examine its use as a heart sounds locator regardless of pulmonary airflow. Lung sounds data was recorded from anterior-right chest locations of six healthy male and female subjects, aged 10-26 years, under three body-mass-standardized flow conditions (low flow: 7.5 mL/s/kg, medium flow: 15 mL/s/kg, high flow: 22.5 mL/s/kg). Suitable window sizes and increment values (chosen based on signal characteristics) were tested, and results are presented for optimal parameters in terms of number of successful heart sound detections. The results show that the VFDT is most successful for heart sound localization at low and medium flow, which are also of most interest. Overall, the method shows promise as a viable technique for this purpose.