WAVELET ANALYSIS OF LUNG SOUNDS

Wavelet Analysis of Lung Sounds

Since there are many questions about the Wavelet analysis of respiratory sound (lung sound), the sample is shown here. Update Dec 24 2000

Normal Vesicular Sounds

The following picture is a image of continuous wavelt transformation of normal vesicular sounds with Gabor8Real as mother wavelet,at resolution level 9 and sublevel 8.
The top stage is a graph of raw normal vesicular sounds, the sampling rate is 22050 Hz. The number of data is 65536.
A wav file of normal vesicular sounds used for this analysis is linked to the following part of Normal Vesicular Sounds.
Normal vesicular sounds can be heard by clicking the link.
Normal Vesicular Sounds This sound kept on record over a left anterior upper lobe of male.

Wheezing

The following sample was recorded over the right anterior upper chest of a boy with asthma.
The following picture is a result of Continuous Wavelet Transformation with Gabor8Real as mother wavelet at resolution level 9 and sublevel 8.
The top stage is a graph of raw Wheezing sounds and sampling rate is 22050Hz. The length of data is 65536.
A wav file of wheezing sounds used for this analysis is linked to the following part of Wheezing.
Wheezing can be heard by clicking the link.
In wheezing, a low peak to the zone of high frequency is seen a little on the left by the center of a screen shown by the arrow . the peak is regarded as peculiar to wheezing ,between relative lebel 22 and 28.
WheezingRecorded over the anterior upper chest of a boy with asthma.

The following picture was obtained by Fast Matching Pursuit with Coiflet4 of the same sound of the bronchial asthma .
The 1024 data is analyzed from viewpoint 24640 or 1.117 seconds of the neighborhood where a characteristic change was seen and which was shown by the arrow in the upper picture.
The wave motion seen on a level 5 is peculiar to asthma. This wave motion shifts to a level 6 and the level 7 with progress of time, and it stops accepting.
This result is in agreement with the picture of the upper continuation wavelet transform.
A level 8 is considered to be the composition ingredient of a normal respiratory sound.
The following picture and the picture under one more was obtained from WaveFMP.exe which was programmed by Hirosi Toda and was published in the "C Magazine" in JAN. 2001.

The following picture was also obtained by Fast Matching Pursuit with Coiflet4 of the same sound of the bronchial asthma .
The 1024 data is analyzed from viewpoint 51520 or 2.335 seconds of the neighborhood where a characteristic change was not seen and which was thought to be normal.
The wave motion seen on the level 5 peculiar to asthma has disappeared.

Crackles

The following picture is a image of Continuous Wavelet Transformation of crackles with Gabor8Real at resolution level 8 and sublevel 8.
Crackles is one of intermittence rale and it is also called putty putty sound.
The sample was Recorded over the right posterior lower chest of a boy with pneumonia.
The top stage is a graph of raw respiratory sounds, the sampling rate is 22050 Hz. The number of data is 32768.
A wav file of clackles sounds used for this analysis is linked to the following part of Clackles.
clackles sounds can be heard by clicking the link.
A little, from the neighborhood of a left arrow, it applies to the neighborhood of a lower right arrow, the peak regarded as peculiar to cracles by lebel20 to lebel46 of a relative level is seen by the center of a screen, and it is gradually observed from it, that frequency is low.
CracklesRecorded over the right posterior lower chest of a boy with pneumonia.

Grunting

The following picture is a result of CWT of Grunting with MorletPower as mother wavelet at lebel 6 and sublevel 16.
Grunting was recorded at the mouth of a premature baby with respiratory distress.
Unlike the above-mentioned data, the number of quantization bits is 16 bits and sampling frequency is 5500Hz.
In the neighborhood shown by the arrow, it turns out that Grunting simultaneously consists of two or three frequency.
The picture which expanded the portion of this arrow further was shown to the next.
In the neighborhood shown by the arrow, it becomes clearer to consist of three frequency.
GruntingRecorded at the mouth of a premature baby with respiratory distress.

The following picture is a image by CWT of the same Grunting data of 1024 from viewpoint 21568 with Morlet Power as mother wavelet.
Maximun resolution level is 5 and Sublevel is 32.
It is observable to have converged, while the low wave motion of frequency and the high wave motion of frequency shown by two arrows occur simultaneously and raise frequency to the usual wave motion of the level 35 neighborhood with progress of time.

The following picture is a image by Wavelet Packet Transformation（WPT) ob grunting form the same drespiratory distress syndrom with Daubechies20 as mother wavelet at resolution level 6.
the data length is 1024 from the viewpoint 21568.
w6,10 to w6,13 crystals are equivalent to the wave motion with the low frequency of the last picture.
w6,16 to w6,22 crystals are equivalent to the wave motion with the high frequency of the last picture.
A high frequency wave motion seems to be divided into two by the last picture.
Also by WPT, the wave motion with low amplitude is seen from w6,23 crystal to w6,25crystal, and is considered to support the view of continuation wavelet transform.

Hereafter, to be continue.

MAIL:wavelet@nifty.ne.jp

Last update: