Twelve subjects were studied. All were wheezing at the time of study and all had moderate-to-severe airways obstruction with an FEV! between 720 ml and 1,500 ml. Table 1 shows the FEV, on pulmonary function tests done before and after bronchodilators, performed within four weeks of the lung sound recording.
Five five-minute periods of observation were studied in each patient. To determine the most efficient number of samples to be randomly chosen, we compared 5, 25, 50, and 75 samples per five-minute period. As shown in Table 2, the ratio for intersubject variation was significantly greater when 50 segments were analyzed rather than five segments. The addition of 25 more segments did not appear to add any additional information. We chose to analyze 50 segments per five-minute period.
Comparing sounds with corresponding spectrums, we found that an audible wheeze was associated with a peak with a narrow frequency band and a frequency between 150 and 1,000 Hz as noted previously. Figure one shows two segments of the sound signal, one with and one without a peak. The FFT of the sound signal associated with a wheeze always had a peak similar to that seen on the left in Figure 1.
The minimum peak height to be considered as a wheeze was studied. Table 3 compares Est Tw/Ttot with FEVb using different criteria for peak height.
Among patients, a significant correlation was found between the FEV! and the percentage of segments with peaks three or more times baseline (Est Tw/Ttot) as shown in Figure 2 (r =.89, p<0.001). Table 4 shows the estimated Tw/Ttot for each subject for all five five-minute segments. All peak heights shown were associated with a significant correlation with FEVb but the use of three times baseline as the minimal peak value was associated with the highest F ratio on analysis of variance comparing intersubject vs intrasubject variation. Asthma sufferers prefer to treat asthma and arrest the attacks with the help of ventolin inhalers.
Five subjects underwent eight sleep studies. None showed significant desaturation during the night. No patient had apnea. Figure 3 shows the sleep stages and wheezing of one patient. The percentage of wheezing (Est Tw/Ttot) for every five-minute period from 11 pm to 6 am is shown. There are several time intervals during the night when there was little wheezing and other times with a large amount of wheezing.
Previous studies have compared airflow obstruction at midnight and at 4 am. The Est Tw/Ttot from midnight to 12:30 am was compared with that from 4:00 to 4:30 am for the eight sleep studies (Fig 4). More wheezing, and therefore more obstruction, was found between 4:00 and 4:30 am (p<0.05).
Figure 1. Frequency spectrum of 250 ms of sound signal. Left panel reveals a sharp peak at 490 Hz corresponding to an audible wheeze. Right panel has no significant frequency peaks above 150 Hz.
Figure 2. Comparison of FEVi to the estimated Tw/Ttot (r = 0.89, p<0.001).
Figure 3. One subject studied through night. Study began at midnight. Lower panel is estimated Tw/Ttot. Upper panel shows corresponding sleep stage (W is awake; R, REM sleep; 1-4, sleep stages 1 to 4).
Figure 4. Est Tw/Ttot for Midnight to 12:30 am vs 4:00 am to 4:30 am (p<0.05).
Table 1—Results of Pulmonary Function Tests Obtained Within Four Weeks of Study
|Subject||Age||FEVj, L||FEV„ L|
Table 2—Accuracy of Estimated TwITtot Using Analysis of Variance: Effect cf Varying the Number of Samples
|No. Samples||Intersubject Variation (F ratio)||Intrasubject Variation (F ratio)|
Table 3—Accuracy af Estimated TwITtot: Effect of Varying the Minimum Peak Height
|Peak Height Compared to Baseline||Correlation Est Tw/Ttot vs FEV,||Intersubject Variation (F Ratio)|
Table 4—Estimated Tw/Ttot for Patients
|FEV,, L||Estimated Tw/Ttot (%)For Each Five Minute Segment|