Sound Recording and Sound Analysis

Figure 1 shows a block diagram of the equipment employed in the lung sound recordings (17). The sound recording system consisted of a handheld assembled microphone unit (prototype device, Omron Healthcare Corporation, Ltd., Kyoto, Japan) and a pulse-code modulation recorder (PCM-D100 series, Sony Corporation, Tokyo, Japan). Two microphones (MP34DR04, STMicroelectronics, Geneva, Switzerland) were set in the microphone unit; one collected ambient sounds around the device, while the other collected lung sounds from the right side of the chest. We resampled the lung sounds at 44.1-kHz with 16-bit quantization, performing a 4,096-point fast Fourier transformation using sound analysis software (Adobe Audition CC 2018, Adobe Inc., San Jose, CA, United States).

Block diagram showing the equipment used for lung sound recording.

Figure 2 shows a sample of the spectrogram and spectrum of the wheezing sounds (17). Prior to the sound analysis, engineering researchers and two trained physicians evaluated all recordings. The recorded lung sounds were reviewed to discriminate wheezes from other sounds, such as noises generated due to friction between the microphone and the participants' skin. We detected wheezing on the sound spectrograms as defined by the Computerized Respiratory Sound Analysis (CORSA) guidelines. A wheeze is defined as a continuous adventitious musical sound. Acoustically, it is characterized by periodic waveforms with a dominant frequency typically over 100 Hz and lasting over 100 ms (18).

Wheezing indexes. The lung sound analysis displayed wheezes as horizontal bars of intensity with corresponding sharp peaks of power. The sharp peak of power was defined as a wheeze. The following indexes were calculated: (1) duration of wheeze power bands (B), (2) number of wheeze bands per 30s (B), (3) lowest frequency of wheeze power bands (B), (4) highest frequency of wheeze power bands (B), and (5) frequency of maximum intensity of wheeze power bands (C).