An analysis of sounds for lungs with excessive water

Excessive lung water occurs when too much water accumulates in the lung, causing breathing difficulty. Current diagnosis methods include X-rays and CT-scans. However, because of their bulk and the need for trained professionals to operate, physicians rely on auscultation for preliminary diagnosis. Recent attempts have been made to automate the auscultation process and some degree of success has been reported. Thus, it would be useful to provide more analysis of such lung sounds. This paper attempts to study the characteristics of breathing sounds from lungs with excessive water and compare them with breathing sounds from healthy lungs. Using a modified empirical mode decomposition to split the signals, the root-mean-squared energy (RMS) and kurtosis were used as characteristics. These characteristics were extracted from the intrinsic mode functions (IMF) and were analyzed. Results showed that certain IMF were effective in characterizing both kinds of sounds due to their small spread in RMS or kurtosis. Results also (using p-values from statistical tests) showed that for certain intrinsic mode functions, lung sounds with excessive lung water exhibit different medians from sounds of healthy lungs. There was strong linear independence between each IMF of the two sounds. Empirical mode decomposition was shown to be able to extract useful information for analyses.