2.3, Breath sound analysis method
Breath sound analyses were performed for all participants, as described
previously13, 16). Breath sounds were recorded using a
handheld microphone for ≥10 seconds. The microphone was placed on the
right upper anterior chest at the second intercostal space along the mid
clavicular line. A sound analysis of the inspiration phase was performed
using an LSA-2012 sound spectrometer (Kenz Medico Co., Saitama, Japan).
The sound-amplifying unit was found to be effective for analyzing sounds
in the range of 100-3,000 Hz. The recorded sounds were analyzed
according to the fast Fourier transformation. The sampling frequency was
10,240 Hz and the spectra were obtained using a Hamming window. The
sounds were displayed as a sound spectrogram. The point of the maximum
frequency (Hz) in the shape during inspiration was used for the sound
spectrum analysis.
To evaluate the dBm-based spectrum images, we used a new analyzing
program (LSA 2020/ANA, Kenz Medico Co., Saitama, Japan). Using this
program, we decided to set the zero point of the Y-axis (dBm) based on
the mean of the background noise of all subjects. In this report, the
zero level (0 dB of breath sound spectrum) was visually corrected based
on the breath sound spectrums in each sample before the zero point (the
frequency at 0 dB) was decided22). The zero level and
the zero point were used to calculate of the area under the curve (AUC)
of the sound spectrum.
From the breath sound spectrum, the power of low frequencies (LF)
(100-200 Hz) and middle frequencies (MF) (200-400 Hz) in the inspiration
and the expiration sounds were calculated17). In
addition, the inspiration-to-expiration sound power ratio in the low-
and mid-frequency ranges (E/I LF and E/I MF, respectively) were also
calculated17).
The sound spectrum curve indices of the inspiration were determined by
one point of the maximum frequency (Hz) in the sound spectrum. The data
were automatically calculated using a custom software
program22). The spectrum curve indices
(A3/AT,B4/AT, RPF75 and
RPF50) were measured according to the methods of a
previous report13, 16). The RPF75 is
the ratio of power at F75 (dBF75) to the
frequency value (F99-F75), and the
RPF50 is the ratio of power at F50(dBF50) to the frequency value
(F99-F50). The total sound spectrum was
divided into three or four sections from low to high frequencies, which
allowed for the quantitation of the energy distribution within the
spectrum13, 16). The AT,
A3 and B4 were conventionally calculated
by dBm and Hz (1 arbitrary unit [dBm・Hz] on a spectrum image. The
A3/AT and
B4/AT values are the ratio of the
higher-frequency areas (A3 and B4) to
the total area (AT) of the sound
spectrum13, 16).
A five-point moving average was used as a smoothing technique to
determine the suitable dB value for identifying the 0 dB in the slope of
each sound spectrum. Each personal breath sound parameter was analyzed
conventionally, using a sample with a median value from three tidal
breaths.