J Appl Physiol 116: 61–66, 2014. First published November 14, 2013; doi:10.1152/japplphysiol.00651.2013.
Changes in breath sound power spectra during experimental oleic acid-induced lung injury in pigs Jukka Räsänen,1 Michael E. Nemergut,2 and Noam Gavriely3 1
Department of Anesthesiology, H. Lee Moffitt Cancer Center, Tampa, Florida; 2Department of Anesthesiology, Mayo Foundation for Education and Research, Rochester, Minnesota; and 3Technion, Israel Institute of Technology, Rappaport Faculty of Medicine, Haifa, Israel Submitted 4 June 2013; accepted in final form 7 November 2013
Räsänen J, Nemergut ME, Gavriely N. Changes in breath sound power spectra during experimental oleic acid-induced lung injury in pigs. J Appl Physiol 116: 61– 66, 2014. First published November 14, 2013; doi:10.1152/japplphysiol.00651.2013.—To evaluate the effect of acute lung injury on the frequency spectra of breath sounds, we made serial acoustic recordings from nondependent, midlung and dependent regions of both lungs in ten 35- to 45-kg anesthetized, intubated, and mechanically ventilated pigs during development of acute lung injury induced with intravenous oleic acid in prone or supine position. Oleic acid injections rapidly produced severe derangements in the gas exchange and mechanical properties of the lung, with an average increase in venous admixture from 16 ⫾ 12 to 62 ⫾ 16% (P ⬍ 0.01), and a reduction in dynamic respiratory system compliance from 25 ⫾ 4 to 14 ⫾ 4 ml/cmH2O (P ⬍ 0.01). A concomitant increase in sound power was seen in all lung regions (P ⬍ 0.05), predominantly in frequencies 150 – 800 Hz. The deterioration in gas exchange and lung mechanics correlated best with concurrent spectral changes in the nondependent lung regions. Acute lung injury increases the power of breath sounds likely secondary to redistribution of ventilation from collapsed to aerated parts of the lung and improved sound transmission in dependent, consolidated areas. acute lung injury; respiratory sounds; acoustics THE DEVELOPMENT OF ACUTE LUNG injury is characterized by loss of alveolar capillary membrane integrity, accumulation of fluid in the extravascular space, and loss of gas volume, particularly in the dependent parts of the lungs (4). The presence of areas of low ventilation-to-perfusion ratios and frank atelectasis or consolidation lead to the clinical manifestations of respiratory failure: arterial hypoxemia and impaired lung mechanics. We have shown previously that these pathophysiological changes during the development of experimental lung injury also increase the transmission amplitude of broad-band sound introduced into the airway (8). When measured at multiple sites around the chest, the increase in sound transmission localizes to areas of lung pathology, being largest in the dependent lung regions where the most extensive lung injury is found both radiographically and at postmortem examination (8, 9). An increase in lung tissue density enhances sound transmission through injured areas of the lung because of better matching between the acoustic impedances of the lung parenchyma and the chest wall. As less sound energy is reflected back at this interface, more sound energy passes through the chest wall to the surface sensor. Previous studies in animals, both with and without experimental lung injury, have also shown that the acoustic spectra of
Address for reprint requests and other correspondence: J. Räsänen, Dept. of Anesthesiology, H. Lee Moffitt Cancer Center, 12902 USF Magnolia Dr., Tampa, FL 33612 (e-mail: [email protected]). http://www.jappl.org
breath sounds produced by mechanical ventilation are altered by changes in lung aeration (5). Attenuation of lung sounds with hyperinflation has been attributed to decreased sound transmission (11), while increases in spectral power in certain frequency bands during experimental lung injury are believed to be caused by appearance of adventitious sounds, primarily crackles (12). We designed this study to investigate potential changes in the spectra of lung sounds, independent of adventitious sounds, and their temporal relationship to changes in gas exchange and lung mechanics during the evolution of experimental lung injury. MATERIALS AND METHODS
After approval by the Institutional Animal Care and Use Committee of the Mayo Clinic, 10 healthy pigs, weighing 35– 45 kg and cared for according to the current guidelines for the care and use of laboratory animals, were included in the study. After an overnight fast, the pigs were anesthetized with an intramuscular injection of 4 mg/kg telazol and 2 mg/kg xylazine. The anesthesia was deepened with inhalation of 3% halothane in oxygen by mask, whereafter the animal was intubated with a 7.5-mm internal diameter tracheal tube. The tracheal tube was connected to an anesthesia circle system, and the animal was mechanically ventilated with 0.5–1.0% isoflurane in air-oxygen mixture using constant inspiratory flow, a tidal volume of 8 –10 ml/kg, ambient end-expiratory airway pressure, and a ventilator rate sufficient to maintain arterial blood carbon dioxide tension within 35– 45 Torr. Only the ventilator rate and the inspired O2 fraction (FIO2) were adjusted as necessary during the study. Tidal volume, airway pressure, and end-tidal carbon dioxide concentration (PETCO2) were monitored at the proximal end of the tracheal tube. A pulse oximeter probe was attached to the animal’s hoof for monitoring heart rate and oxyhemoglobin saturation (SpO2). A femoral artery was cannulated for direct measurement of blood pressure and for sampling of arterial blood. A pulmonary artery catheter was inserted via the right jugular vein for monitoring of central venous, pulmonary artery and pulmonary capillary occlusion pressures, mixed-venous SpO2, cardiac output, core temperature, and for sampling of mixed-venous blood. Throughout the study, Ringer lactate was infused to maintain pulmonary artery and pulmonary capillary occlusion pressure between 9 and 14 mmHg; core temperature was maintained with a heating blanket. Breath sounds were recorded using six PPG sensors (Linear 50 – 2,000 Hz, response range 10 –5,000 Hz, PPG-02; Technion, Haifa, Israel) placed on the chest wall. The six sensors were positioned three over each lung in the following locations: 1) posteriorly 5 cm lateral from the spine, 2) anteriorly 5 cm lateral from the edge of the sternum, and 3) on the side, halfway between the other two locations. In the cranio-caudal direction, the sensors were halfway between the apex and base of the lung. The sensors were secured to the shaved chest with a circumferential elastic strip, such that firm contact was maintained, but inspiratory chest expansion was unimpeded, without excessive pressure on the sensors. Each sound recording was of 10-s
8750-7587/14 Copyright © 2014 the American Physiological Society Downloaded from www.physiology.org/journal/jappl (109.236.055.194) on February 13, 2019.
61
62
Breath Sound Spectra in Experimental Lung Injury
•
Räsänen J et al.
Fig. 1. A section of a recording during lung injury with inspiratory breath sounds in waveform and spectral display. The timing of the 150-ms selection is marked showing its relationship to end-inspiratory crackles (arrows). A highpass filter at 150 Hz is applied to attenuate heart sounds.
compliance value was averaged over 3 min, coinciding with the time of each sound recording during development of the injury. The sound recordings for each sensor and each phase of the study were analyzed manually using sound-editing software (Adobe Audition CS6, Adobe, San Jose, CA). The beginning of each inspiration was identified from a spectral display, and a 150-ms segment was selected starting 50 ms into the inspiration (Fig. 1). Visual and auditory examination of the segment was used to ascertain that no adventitious sounds were included in the selection. If such sounds were present, the selection was reduced in length to 100 ms to exclude the adventitious sound. The selected segment was scanned, and a 256-point fast Fourier transformation (FFT) was generated using a Blackmann-Harris window. The FFT results of the three consecutive inspirations in each recording were averaged at each frequency point to yield the final power spectrum for a given recording (Fig. 2). Both of the duplicate recordings were analyzed and averaged for each phase of the study. A band of 150 –5,000 Hz of the frequency spectrum was isolated for analysis. The band was subdivided into three additional subbands (150 – 800; 800 –3,000, and 3,000 –5,000 Hz). The spectral power within each band was calculated as an average of the FFT points belonging to the band.
-40 -50
Power (dB)
duration and contained three respiratory cycles. The signal from each sensor was amplified (Blue tube, Presonus, Baton Rouge, LA) and digitized (10,000 samples/s, 16-bit) into a portable computer using a 16-channel, 16-bit analog/digital converter board (PCI-6035E, National Instruments, Austin, TX) inserted into a PCI expansion system (CB1F, Magma, San Diego, CA). All sound measurements were made in duplicate, with the second recording immediately following the first. Past experiments have shown that the extent of aeration during this type of lung injury is influenced by gravity, with the dependent lung regions being the most severely affected (8, 11). To ensure that any observed differences between lung regions were related to gravity rather than the anatomical sites themselves, we assigned each animal randomly to have the experiment conducted in either supine or prone position. Experimental protocol. After instrumentation, the animals were turned to their randomly selected starting position, prone or supine, and allowed to stabilize for 15 min. After baseline respiratory volume and pressure recordings, blood-gas sampling, hemodynamic measurements, and duplicate sound recordings, acute lung injury was induced with oleic acid, as previously described (8). FIO2 was adjusted to maintain arterial O2 saturation ⬎ 90% and respiratory rate to keep PETCO2 ⬍ 50 Torr. After the oleic acid injection, single-sound recordings were made at 5-min intervals as the lung injury progressed. At the same time, arterial and mixed-venous blood-gas samples were collected every 10 min, and cardiopulmonary monitoring data were automatically saved at 1-min intervals. The lung injury was considered fully developed when the deteriorating trend in oxygenation and lung mechanics leveled off. At that time, duplicate sound samples and blood-gas values were again recorded, after which the study was concluded, and the animal was euthanized. A postmortem examination of the lungs was performed with the animal in the position assigned by randomization. Data analysis. Venous admixture was calculated using the standard formula: dynamic respiratory system compliance was calculated by dividing the tidal airway pressure change by the expiratory tidal volume. In addition, minute-to-minute values for dynamic lung compliance, FIO2, SpO2, and venous O2 saturation were obtained from the monitor data file. The ventilation-perfusion index was calculated from the FIO2 and saturation values as an estimate of venous admixture, as described previously (6, 7). Each ventilation-perfusion index and
-60 -70 -80 -90
0
1000
2000
3000
4000
5000
Frequency (Hz) Fig. 2. Power spectra from the 150-ms early inspiratory selections of three consecutive breaths (shaded lines) and their average (black line) from a single recording in one anesthetized, mechanically ventilated pig.
J Appl Physiol • doi:10.1152/japplphysiol.00651.2013 • www.jappl.org Downloaded from www.physiology.org/journal/jappl (109.236.055.194) on February 13, 2019.
Breath Sound Spectra in Experimental Lung Injury
RESULTS
Nine animals, five studied supine and four prone, survived the sequence of interventions; one animal died of acute pulmonary hypertension and intractable right heart failure soon after oleic acid injection. The average time from the oleic acid injection to fully developed injury was 53 ⫾ 17 min (range 15–75 min). In one animal randomized to start in the prone position, the injury developed too rapidly (15 min) to provide a sufficient number of data points for the injury development phase. Data from this animal is, therefore, used only to compare the full effect of injury. In all other animals, at least 50 min of data were available. Physiological measurements. The sequence in which the prone/supine positioning was implemented did not have a statistically significant effect on any of the measured physiological variables. Hence data could be averaged across the two groups of animals studied prone and supine. Induction of lung injury was associated with statistically significant pulmonary hypertension, without a significant change in mean systemic arterial pressure, heart rate, or cardiac output (Table 1). Venous admixture increased fourfold, and oxygen supplementation was required to prevent profound hypoxemia in all subjects. The dynamic respiratory system compliance decreased to 43% of its baseline. Decreased tidal alveolar ventilation necessitated an increase in ventilator rate to prevent a rise in PETCO2 and arterial blood carbon dioxide tension beyond the planned upper limit. The blood hemoglobin concentration increased slightly, but statistically significantly, as injury developed. Table 1. The effects of oleic acid-induced lung injury on cardiopulmonary function in 9 pigs Heart rate, beats/min Mean blood pressure, mmHg Mean PA pressure, mmHg CVP, mmHg Cardiac output, l/min Core temperature, °C Hemoglobin, g/l Ventilator rate, cpm Tidal volume, ml F IO2 SaO2, % PaCO2, Torr pHa Venous admixture, % Cdyn, ml/cmH2O
Baseline
Injury
P Value
100 ⫾ 14 89 ⫾ 13 23 ⫾ 4 5⫾5 4.4 ⫾ 1.3 36.1 ⫾ 0.9 11.2 ⫾ 0.9 18 ⫾ 2 349 ⫾ 44 0.38 ⫾ 0.07 97.0 ⫾ 0.6 43 ⫾ 4 7.40 ⫾ 0.06 16 ⫾ 12 25 ⫾ 4
111 ⫾ 18 80 ⫾ 14 35 ⫾ 6 8⫾5 4.0 ⫾ 1.1 36.2 ⫾ 1.0 13.0 ⫾ 1.7 19 ⫾ 2 350 ⫾ 52 0.75 ⫾ 0.25 89.4 ⫾ 8.0 60 ⫾ 6 7.23 ⫾ 0.06 62 ⫾ 16 14 ⫾ 4
NS NS ⬍0.01 NS NS NS ⬍0.05 NS NS ⬍0.05 ⬍0.05 ⬍0.01 ⬍0.01 ⬍0.01 ⬍0.01
Values are means ⫾ SD. PA, pulmonary artery; CVP, central venous pressure; cpm, cycles per minute; FIO2, inspired O2 fraction; SaO2, arterial O2 saturation; PaCO2, arterial PCO2; pHa, arterial pH; Cdyn, dynamic respiratory system compliance; NS, nonsignificant.
63
Räsänen J et al.
Table 2. The effects of oleic acid-induced lung injury on average acoustic power over the measurement range of 150⫺5,000 Hz, broken down by side of recording and the position of the animal during the experiment Average Power
Total Left Right Prone Supine
Baseline
Injury
P Value
⫺73.9 ⫾ 1.8 ⫺73.2 ⫾ 2.3 ⫺73.9 ⫾ 2.3 ⫺74.6 ⫾ 1.6 ⫺73.3 ⫾ 1.9
⫺71.9 ⫾ 1.9 ⫺72.6 ⫾ 1.4 ⫺72.3 ⫾ 1.9 ⫺72.0 ⫾ 2.8 ⫺71.8 ⫾ 1.0
⬍0.01 NS NS NS NS
Values are means ⫾ SD in dB.
In all cases, examination of the lungs at the end of the study revealed consolidation of the dependent part of both lungs over one-fifth to one-third of their vertical height. This finding was consistent, and its anatomical distribution reflected the randomly assigned position, prone or supine, in which the animal finished the experiment. Sound measurements. Oleic acid lung injury effected a statistically significant increase in the average spectral power of breath sounds when calculated over the entire measurement band across all six sensors (Table 2; P ⬍ 0.01). Averaging the power from the three sensors overlying each lung or over the animals studied prone or supine did not reveal statistically significant differences, or a statistically significant interaction with the effect of the injury itself. Hence, in subsequent analyses of the effect of gravity and different frequency bands, data could be averaged across both lungs and all animals. Breath sound power spectra recorded during the development of lung injury over the right dependent lung region in a representative animal are shown in Fig. 3. When the change in sound power was plotted over the studied frequency range, we noted three frequency bands in which the effect of injury appeared to be distinctly different (Fig. 4). The increase was largest within 150 – 800 Hz, moderate within 800 –3,000 Hz, and negligible at frequencies exceeding 3,000 Hz. This frequency dependency of the power change was statistically significant (P ⬍ 0.001). We also found that the frequency distribution of this power change was different for the three lung regions (Figs. 4 and 5). The increase -40 -50
Power (dB)
The results are presented as means ⫾ SD. The Wilcoxon signedrank test or Friedman’s nonparametric repeated-measures analysis of variance, where appropriate, were used to test the effects of lung injury on the measured variables. If the analysis of variance showed statistical significance, relevant individual differences between groups were tested with Dunn’s multiple-comparison test. The strength of association between changes in spectral power and physiological variables reflecting lung injury was tested with linear regression analysis performed on population averages. Results were considered statistically significant if the probability of type ␣ error was ⬍5%.
•
-60 -70 -80 -90
0
1000
2000
3000
4000
5000
Frequency (Hz) Fig. 3. Power spectra of inspiratory breath sounds recorded over the dependent portion of the right lung in one anesthetized, mechanically ventilated pig during development of oleic acid-induced lung injury. Dashed black line represents sound power distribution during baseline conditions, and solid black line during fully developed injury. The thin shaded lines represent intermediate recordings as the injury develops.
J Appl Physiol • doi:10.1152/japplphysiol.00651.2013 • www.jappl.org Downloaded from www.physiology.org/journal/jappl (109.236.055.194) on February 13, 2019.
64
Breath Sound Spectra in Experimental Lung Injury
8
8
Power (dB)
10
Power (dB)
10
6 4 2 0
•
Räsänen J et al. Supine Anterior Supine Posterior Prone Anterior
6
Prone Posterior
4 2 0
1000
-2
2000
3000
4000
5000 -2
Frequency (Hz)
1000
2000
3000
4000
5000
Frequency (Hz)
Fig. 4. Changes (⌬) in breath sound power at different frequencies resulting from oleic acid-induced lung injury recorded over nondependent (dotted line), midlung (dashed line), and dependent (solid line) locations in nine anesthetized, mechanically ventilated pigs. The response of spectral power change to injury is distinctly different in three frequency regions with boundaries at 800 and 3,000 Hz.
Fig. 6. Changes in breath sound power at different frequencies resulting from oleic acid-induced lung injury recorded over anterior and posterior chest regions in 5 supine and 4 prone, mechanically ventilated pigs. A similar change, independent of sensor location and position, is seen at a frequency band of 150 – 800 Hz. At higher frequencies, a change occurs primarily in the dependent parts of the lung, regardless of position.
was similarly large in all regions at low frequencies from 150 to 800 Hz. In the nondependent and midlung regions, the change in power was statistically significantly smaller in the 800- to 3,000-Hz (P ⬍ 0.05) and 3,000- to 5,000-Hz (P ⬍ 0.001) bands than it was at lower frequencies. In the dependent lung regions, the change was equally large in the 150- to 800-Hz and the 800- to 3,000-Hz bands, and only decreased at frequencies over 3,000 Hz (P ⬍ 0.05). Separate plots of the change in sound power in the prone and supine positions (Fig. 6) illustrate further that the increase observed in the 150- to 800-Hz range occurred in all lung areas, whereas the increase in the 800- to 3,000-Hz band developed always in the dependent areas, anterior or posterior, depending on the subject’s position prone or supine, respectively. In eight of the nine animals, we were able to compare the changes in spectral power with concurrent changes in oxygenation and lung mechanics as the lung injury was developing. The linear correlation coefficients between these variables and the three gravity-related lung regions are shown in Table 3 for two frequency bands (150 – 800 Hz, 800 –3,000 Hz). The correlation coefficients were highest for the spectral power recorded over the nondependent lung regions at the low-frequency range, and somewhat higher
for oxygenation than for lung mechanics. The time course of spectral power changes, oxygenation, and lung mechanics showed similar progression in each of the variables, starting from the first measurement 5 min after the injection of oleic acid (Fig. 7).
10
Power (dB)
8 6
*
4
*
*
2 0
DISCUSSION
Although the frequency spectra of breath sounds during spontaneous breathing in humans with normal lungs and chronic pulmonary disease have been well characterized (2), little information is available about spectral changes during mechanical ventilation and acute lung injury. Ploysongsang et al. (5) studied inspiratory frequency spectra from dogs during development of hydrostatic pulmonary edema and reported increases in mean, median, and mode frequencies that preceded the development of adventitious sounds. Vena et al. (11) recorded inspiratory breath sound spectra from healthy pigs during application of progressively higher levels of positive end-expiratory pressure and reported an attenuation and downward shifting of sound spectra that correlated with end-expiratory lung volume. In a subsequent study, Vena et al. (12) effected unilateral lung injury in pigs and recorded the appearance of pathological spectral components in a frequency range above 500 Hz in the injured lung. These changes were attributed to the appearance of crackles, they correlated with de-aeration in lung tissue, and were reversible with the application of positive end-expiraTable 3. Linear correlation coefficients (r2) between acoustic power and measures of oxygenation (VQI) and lung mechanics (Cdyn) during development of oleic acid-induced lung injury in eight pigs Frequency/Location
150-800
800-3000
3000-5000
Frequency (Hz) Fig. 5. Average differences in breath sound spectral power in three gravitational locations and three frequency bands in nine pigs between baseline conditions and oleic acid-induced lung injury. Values are means ⫾ SD. Open bars, nondependent lung; shaded bars, midlung; solid bars, dependent lung. *P ⬍ 0.05, †P ⬍ 0.001 compared with 150 – 800 Hz within the same lung region.
150–800 Hz Nondependent Midlung Dependent 800–3,000 Hz Nondependent Midlung Dependent
VQI, %
Cdyn, ml/cmH2O
0.86 0.84 0.71
0.76 0.75 0.60
0.18 0.22 0.36
0.53 0.29 0.36
VQI, ventilation-perfusion index.
J Appl Physiol • doi:10.1152/japplphysiol.00651.2013 • www.jappl.org Downloaded from www.physiology.org/journal/jappl (109.236.055.194) on February 13, 2019.
Breath Sound Spectra in Experimental Lung Injury
6
20 4
15 10
2
Power (dB)
VQI (%) Cdyn (ml/cmH20)
25
5 0
0
5 10 15 20 25 30 35 40 45 50
0
Time from injection (min) Fig. 7. Time course of changes in spectral power over nondependent lung regions (shaded bars), oxygenation [ventilation-perfusion index (VQI), solid bars], and lung mechanics [dynamic respiratory system compliance (Cdyn), open bars] during development of oleic acid-induced lung injury in eight pigs. Time zero represents baseline. Breath sound power is normalized to baseline value.
tory pressure. In both of these studies, the entire inspiratory phase of the breath was included in the spectral analysis, and no attempt was made to separate out breath sounds generated by turbulent gas flow and adventitious sounds. We designed the present study to determine whether the spectral characteristics of the breath sounds themselves, isolated from adventitious sounds, would provide an indication of injury. Such an analysis was feasible because adventitious sounds appear late in inspiration during oleic acid-induced lung injury, leaving a clean inspiratory spectral sample earlier in the inspiratory phase (Fig. 1). Monitoring of sound spectra might then provide a noninvasive means of detecting, localizing, and monitoring lung injury with an array of acoustic sensors. The results of our study revealed that oleic acid-induced lung injury induces a readily detectable increase in spectral power throughout the lung, and that this effect was commensurate with conventional measures of lung injury. Breath sounds are generated in the airways by turbulent flow, and their intensity is proportional to the gas flow rate in the airways (1). The sound waves then must propagate through the lung parenchyma, crossing the pleural space and chest wall to allow detection by a sensor on the outside surface. An increase in the power of the recorded signal implies that its original intensity has changed, the characteristics of its propagation through the respiratory system have been altered, or a new, additional signal has been added. All of these mechanisms can conceivably operate in injured lungs. Oleic acid-induced experimental lung injury disrupts the alveolar capillary membrane and causes volume loss in the injured lung (3, 10). The degree of volume loss is modified by gravity, such that the dependent lung is atelectatic and receives no gas flow, while the nondependent parts have relatively normal patency; the intermediate areas constitute a transitional spectrum of abnormalities between the two extremes (8, 9, 12). If under these circumstances the flow rate of tidal ventilation is maintained unchanged, as it was in the present study, then the gas flow rate in the airways that still conduct ventilation will become higher. Breath sounds generated in these airways will be intensified due to the
•
Räsänen J et al.
65
higher Reynolds numbers. We speculate that this is the primary mechanism for the increased acoustic power observed in the nondependent and perhaps in the midlung areas in this study. If redistribution of inspiratory gas flow were the only mechanism affecting acoustic power, then a decrease in spectral power would be expected over the dependent areas, as inspiratory flow is diverted away from them. This we did not observe; on the contrary, spectral power increase appeared to be more sustained and spread from the 150- to 800-Hz frequency range into the 800- to 3,000-Hz band in those areas. Since adventitious sounds from the spectral samples were meticulously excluded, there must have been an enhancement of sound propagation in the dependent areas to explain the increased spectral power, despite reduced air entry. We have previously shown that sound transmission through the lung parenchyma is indeed enhanced by injury in oleic acid-induced pulmonary edema in proportion to the gravity-associated involvement of any given lung region (9). This effect is likely secondary to the reduction of acoustic impedance at the lung-chest wall interface, as lung parenchyma loses gas volume. Even though the studies proving this were conducted using an external signal introduced into the airway, the same phenomenon would be expected to facilitate the conduction of natural breath sounds as well. The increase in spectral power, particularly in the nondependent areas of the lung, correlated well with conventional measures of oxygenation and lung mechanics. This is not surprising, as changes in both venous admixture and respiratory system compliance during lung injury are secondary to the development of lung regions with zero or low ventilation-to-perfusion ratios. The fact that changes in these three variables originate from the same underlying pathology indicates that their response to the progression or regression of injury will be similar, which indeed was evident in our study. Thus the monitoring of spectral changes would not be expected to provide a time advantage over monitoring oxygenation or lung mechanics. Compared with other acoustic characteristics, such as adventitious sounds, however, the early inspiratory spectral changes do occur earlier in our and others’ experience (5, 11). Our results show that early inspiratory spectral power of breath sounds increases during experimental oleic acidinduced lung injury at a rate commensurate with deterioration of oxygenation and respiratory system mechanics. Since these changes can be detected in all areas of the lung, we postulate a dual mechanism of inspiratory flow redistribution and enhanced propagation of acoustic energy to explain our observations. GRANTS This work was supported by a CR 20 grant from the Mayo Foundation. DISCLOSURES No conflicts of interest, financial or otherwise, are declared by the author(s). AUTHOR CONTRIBUTIONS Author contributions: J.R. and N.G. conception and design of research; J.R. and N.G. performed experiments; J.R., M.E.N., and N.G. analyzed data; J.R., M.E.N., and N.G. interpreted results of experiments; J.R. prepared figures; J.R. drafted
J Appl Physiol • doi:10.1152/japplphysiol.00651.2013 • www.jappl.org Downloaded from www.physiology.org/journal/jappl (109.236.055.194) on February 13, 2019.
66
Breath Sound Spectra in Experimental Lung Injury
manuscript; J.R., M.E.N., and N.G. edited and revised manuscript; J.R., M.E.N., and N.G. approved final version of manuscript. REFERENCES 1. Gavriely N, Nissan M, Rubin AHE, Cugell DW. Spectral characteristics of chest-wall breath sounds in normal subjects. Thorax 50: 1292–1300, 1995. 2. Gavriely N, Cugell DW. Airflow, amplitude and spectral content of normal breath sounds. J Appl Physiol 80: 5–13, 1996. 3. Neumann P, Hedenstierna G. Ventilation-perfusion distributions in different porcine lung injury models. Acta Anaesthesiol Scand 45: 78 –86, 2001. 4. Pelosi P, D’Andrea L, Vitale G, Pesenti A, Gattinoni L. Vertical gradient of regional lung inflation in adult respiratory distress syndrome. Am J Respir Crit Care Med 149: 8 –13, 1994. 5. Ploysongsang Y, Michel RP, Rossi A, Zocchi L, Milic-Emili J, Staub NC. Early detection of pulmonary congestion and edema in dogs by using lung sounds. J Appl Physiol 66: 2061–2070, 1989. 6. Räsänen J, Downs JB, Hodges MR. Continuous monitoring of gas exchange and oxygen use with dual oximetry. J Clin Anesth 1: 3–8, 1988.
•
Räsänen J et al.
7. Räsänen J, Downs JB, Malec DJ, DeHaven B, Garner WL. Real-time continuous estimation of gas exchange by dual oximetry. Intensive Care Med 14: 118 –122, 1988. 8. Räsänen J, Gavriely N. Detection of porcine oleic acid-induced acute lung injury using pulmonary acoustics. J Appl Physiol 93: 51–57, 2002. 9. Räsänen J, Gavriely N. Response of acoustic transmission to positive airway pressure therapy in experimental lung injury. Intensive Care Med 31: 1434 –1441, 2005. 10. Rosenthal C, Caronia C, Quinn C, Lugo N, Sagy M. A comparison among animal models of acute lung injury. Crit Care Med 26: 912–916, 1998. 11. Vena A, Perchiazzi G, Giuliani R, Fiore T, Hedenstierna G. Acoustic effects of positive end-expiratory pressure on normal lung sounds in mechanically ventilated pigs. Clin Physiol Funct Imaging 26: 45–53, 2006. 12. Vena A, Rylander C, Perchiazzi G, Giuliani R, Hedenstierna G. Lung sound analysis correlates to injury and recruitment as identified by computed tomography: an experimental study. Intensive Care Med 37: 1378 – 1383, 2011.
J Appl Physiol • doi:10.1152/japplphysiol.00651.2013 • www.jappl.org Downloaded from www.physiology.org/journal/jappl (109.236.055.194) on February 13, 2019.
Changes in breath sound power spectra during experimental oleic acid-induced lung injury in pigs Jukka Räsänen,1 Michael E. Nemergut,2 and Noam Gavriely3 1
Department of Anesthesiology, H. Lee Moffitt Cancer Center, Tampa, Florida; 2Department of Anesthesiology, Mayo Foundation for Education and Research, Rochester, Minnesota; and 3Technion, Israel Institute of Technology, Rappaport Faculty of Medicine, Haifa, Israel Submitted 4 June 2013; accepted in final form 7 November 2013
Räsänen J, Nemergut ME, Gavriely N. Changes in breath sound power spectra during experimental oleic acid-induced lung injury in pigs. J Appl Physiol 116: 61– 66, 2014. First published November 14, 2013; doi:10.1152/japplphysiol.00651.2013.—To evaluate the effect of acute lung injury on the frequency spectra of breath sounds, we made serial acoustic recordings from nondependent, midlung and dependent regions of both lungs in ten 35- to 45-kg anesthetized, intubated, and mechanically ventilated pigs during development of acute lung injury induced with intravenous oleic acid in prone or supine position. Oleic acid injections rapidly produced severe derangements in the gas exchange and mechanical properties of the lung, with an average increase in venous admixture from 16 ⫾ 12 to 62 ⫾ 16% (P ⬍ 0.01), and a reduction in dynamic respiratory system compliance from 25 ⫾ 4 to 14 ⫾ 4 ml/cmH2O (P ⬍ 0.01). A concomitant increase in sound power was seen in all lung regions (P ⬍ 0.05), predominantly in frequencies 150 – 800 Hz. The deterioration in gas exchange and lung mechanics correlated best with concurrent spectral changes in the nondependent lung regions. Acute lung injury increases the power of breath sounds likely secondary to redistribution of ventilation from collapsed to aerated parts of the lung and improved sound transmission in dependent, consolidated areas. acute lung injury; respiratory sounds; acoustics THE DEVELOPMENT OF ACUTE LUNG injury is characterized by loss of alveolar capillary membrane integrity, accumulation of fluid in the extravascular space, and loss of gas volume, particularly in the dependent parts of the lungs (4). The presence of areas of low ventilation-to-perfusion ratios and frank atelectasis or consolidation lead to the clinical manifestations of respiratory failure: arterial hypoxemia and impaired lung mechanics. We have shown previously that these pathophysiological changes during the development of experimental lung injury also increase the transmission amplitude of broad-band sound introduced into the airway (8). When measured at multiple sites around the chest, the increase in sound transmission localizes to areas of lung pathology, being largest in the dependent lung regions where the most extensive lung injury is found both radiographically and at postmortem examination (8, 9). An increase in lung tissue density enhances sound transmission through injured areas of the lung because of better matching between the acoustic impedances of the lung parenchyma and the chest wall. As less sound energy is reflected back at this interface, more sound energy passes through the chest wall to the surface sensor. Previous studies in animals, both with and without experimental lung injury, have also shown that the acoustic spectra of
Address for reprint requests and other correspondence: J. Räsänen, Dept. of Anesthesiology, H. Lee Moffitt Cancer Center, 12902 USF Magnolia Dr., Tampa, FL 33612 (e-mail: [email protected]). http://www.jappl.org
breath sounds produced by mechanical ventilation are altered by changes in lung aeration (5). Attenuation of lung sounds with hyperinflation has been attributed to decreased sound transmission (11), while increases in spectral power in certain frequency bands during experimental lung injury are believed to be caused by appearance of adventitious sounds, primarily crackles (12). We designed this study to investigate potential changes in the spectra of lung sounds, independent of adventitious sounds, and their temporal relationship to changes in gas exchange and lung mechanics during the evolution of experimental lung injury. MATERIALS AND METHODS
After approval by the Institutional Animal Care and Use Committee of the Mayo Clinic, 10 healthy pigs, weighing 35– 45 kg and cared for according to the current guidelines for the care and use of laboratory animals, were included in the study. After an overnight fast, the pigs were anesthetized with an intramuscular injection of 4 mg/kg telazol and 2 mg/kg xylazine. The anesthesia was deepened with inhalation of 3% halothane in oxygen by mask, whereafter the animal was intubated with a 7.5-mm internal diameter tracheal tube. The tracheal tube was connected to an anesthesia circle system, and the animal was mechanically ventilated with 0.5–1.0% isoflurane in air-oxygen mixture using constant inspiratory flow, a tidal volume of 8 –10 ml/kg, ambient end-expiratory airway pressure, and a ventilator rate sufficient to maintain arterial blood carbon dioxide tension within 35– 45 Torr. Only the ventilator rate and the inspired O2 fraction (FIO2) were adjusted as necessary during the study. Tidal volume, airway pressure, and end-tidal carbon dioxide concentration (PETCO2) were monitored at the proximal end of the tracheal tube. A pulse oximeter probe was attached to the animal’s hoof for monitoring heart rate and oxyhemoglobin saturation (SpO2). A femoral artery was cannulated for direct measurement of blood pressure and for sampling of arterial blood. A pulmonary artery catheter was inserted via the right jugular vein for monitoring of central venous, pulmonary artery and pulmonary capillary occlusion pressures, mixed-venous SpO2, cardiac output, core temperature, and for sampling of mixed-venous blood. Throughout the study, Ringer lactate was infused to maintain pulmonary artery and pulmonary capillary occlusion pressure between 9 and 14 mmHg; core temperature was maintained with a heating blanket. Breath sounds were recorded using six PPG sensors (Linear 50 – 2,000 Hz, response range 10 –5,000 Hz, PPG-02; Technion, Haifa, Israel) placed on the chest wall. The six sensors were positioned three over each lung in the following locations: 1) posteriorly 5 cm lateral from the spine, 2) anteriorly 5 cm lateral from the edge of the sternum, and 3) on the side, halfway between the other two locations. In the cranio-caudal direction, the sensors were halfway between the apex and base of the lung. The sensors were secured to the shaved chest with a circumferential elastic strip, such that firm contact was maintained, but inspiratory chest expansion was unimpeded, without excessive pressure on the sensors. Each sound recording was of 10-s
8750-7587/14 Copyright © 2014 the American Physiological Society Downloaded from www.physiology.org/journal/jappl (109.236.055.194) on February 13, 2019.
61
62
Breath Sound Spectra in Experimental Lung Injury
•
Räsänen J et al.
Fig. 1. A section of a recording during lung injury with inspiratory breath sounds in waveform and spectral display. The timing of the 150-ms selection is marked showing its relationship to end-inspiratory crackles (arrows). A highpass filter at 150 Hz is applied to attenuate heart sounds.
compliance value was averaged over 3 min, coinciding with the time of each sound recording during development of the injury. The sound recordings for each sensor and each phase of the study were analyzed manually using sound-editing software (Adobe Audition CS6, Adobe, San Jose, CA). The beginning of each inspiration was identified from a spectral display, and a 150-ms segment was selected starting 50 ms into the inspiration (Fig. 1). Visual and auditory examination of the segment was used to ascertain that no adventitious sounds were included in the selection. If such sounds were present, the selection was reduced in length to 100 ms to exclude the adventitious sound. The selected segment was scanned, and a 256-point fast Fourier transformation (FFT) was generated using a Blackmann-Harris window. The FFT results of the three consecutive inspirations in each recording were averaged at each frequency point to yield the final power spectrum for a given recording (Fig. 2). Both of the duplicate recordings were analyzed and averaged for each phase of the study. A band of 150 –5,000 Hz of the frequency spectrum was isolated for analysis. The band was subdivided into three additional subbands (150 – 800; 800 –3,000, and 3,000 –5,000 Hz). The spectral power within each band was calculated as an average of the FFT points belonging to the band.
-40 -50
Power (dB)
duration and contained three respiratory cycles. The signal from each sensor was amplified (Blue tube, Presonus, Baton Rouge, LA) and digitized (10,000 samples/s, 16-bit) into a portable computer using a 16-channel, 16-bit analog/digital converter board (PCI-6035E, National Instruments, Austin, TX) inserted into a PCI expansion system (CB1F, Magma, San Diego, CA). All sound measurements were made in duplicate, with the second recording immediately following the first. Past experiments have shown that the extent of aeration during this type of lung injury is influenced by gravity, with the dependent lung regions being the most severely affected (8, 11). To ensure that any observed differences between lung regions were related to gravity rather than the anatomical sites themselves, we assigned each animal randomly to have the experiment conducted in either supine or prone position. Experimental protocol. After instrumentation, the animals were turned to their randomly selected starting position, prone or supine, and allowed to stabilize for 15 min. After baseline respiratory volume and pressure recordings, blood-gas sampling, hemodynamic measurements, and duplicate sound recordings, acute lung injury was induced with oleic acid, as previously described (8). FIO2 was adjusted to maintain arterial O2 saturation ⬎ 90% and respiratory rate to keep PETCO2 ⬍ 50 Torr. After the oleic acid injection, single-sound recordings were made at 5-min intervals as the lung injury progressed. At the same time, arterial and mixed-venous blood-gas samples were collected every 10 min, and cardiopulmonary monitoring data were automatically saved at 1-min intervals. The lung injury was considered fully developed when the deteriorating trend in oxygenation and lung mechanics leveled off. At that time, duplicate sound samples and blood-gas values were again recorded, after which the study was concluded, and the animal was euthanized. A postmortem examination of the lungs was performed with the animal in the position assigned by randomization. Data analysis. Venous admixture was calculated using the standard formula: dynamic respiratory system compliance was calculated by dividing the tidal airway pressure change by the expiratory tidal volume. In addition, minute-to-minute values for dynamic lung compliance, FIO2, SpO2, and venous O2 saturation were obtained from the monitor data file. The ventilation-perfusion index was calculated from the FIO2 and saturation values as an estimate of venous admixture, as described previously (6, 7). Each ventilation-perfusion index and
-60 -70 -80 -90
0
1000
2000
3000
4000
5000
Frequency (Hz) Fig. 2. Power spectra from the 150-ms early inspiratory selections of three consecutive breaths (shaded lines) and their average (black line) from a single recording in one anesthetized, mechanically ventilated pig.
J Appl Physiol • doi:10.1152/japplphysiol.00651.2013 • www.jappl.org Downloaded from www.physiology.org/journal/jappl (109.236.055.194) on February 13, 2019.
Breath Sound Spectra in Experimental Lung Injury
RESULTS
Nine animals, five studied supine and four prone, survived the sequence of interventions; one animal died of acute pulmonary hypertension and intractable right heart failure soon after oleic acid injection. The average time from the oleic acid injection to fully developed injury was 53 ⫾ 17 min (range 15–75 min). In one animal randomized to start in the prone position, the injury developed too rapidly (15 min) to provide a sufficient number of data points for the injury development phase. Data from this animal is, therefore, used only to compare the full effect of injury. In all other animals, at least 50 min of data were available. Physiological measurements. The sequence in which the prone/supine positioning was implemented did not have a statistically significant effect on any of the measured physiological variables. Hence data could be averaged across the two groups of animals studied prone and supine. Induction of lung injury was associated with statistically significant pulmonary hypertension, without a significant change in mean systemic arterial pressure, heart rate, or cardiac output (Table 1). Venous admixture increased fourfold, and oxygen supplementation was required to prevent profound hypoxemia in all subjects. The dynamic respiratory system compliance decreased to 43% of its baseline. Decreased tidal alveolar ventilation necessitated an increase in ventilator rate to prevent a rise in PETCO2 and arterial blood carbon dioxide tension beyond the planned upper limit. The blood hemoglobin concentration increased slightly, but statistically significantly, as injury developed. Table 1. The effects of oleic acid-induced lung injury on cardiopulmonary function in 9 pigs Heart rate, beats/min Mean blood pressure, mmHg Mean PA pressure, mmHg CVP, mmHg Cardiac output, l/min Core temperature, °C Hemoglobin, g/l Ventilator rate, cpm Tidal volume, ml F IO2 SaO2, % PaCO2, Torr pHa Venous admixture, % Cdyn, ml/cmH2O
Baseline
Injury
P Value
100 ⫾ 14 89 ⫾ 13 23 ⫾ 4 5⫾5 4.4 ⫾ 1.3 36.1 ⫾ 0.9 11.2 ⫾ 0.9 18 ⫾ 2 349 ⫾ 44 0.38 ⫾ 0.07 97.0 ⫾ 0.6 43 ⫾ 4 7.40 ⫾ 0.06 16 ⫾ 12 25 ⫾ 4
111 ⫾ 18 80 ⫾ 14 35 ⫾ 6 8⫾5 4.0 ⫾ 1.1 36.2 ⫾ 1.0 13.0 ⫾ 1.7 19 ⫾ 2 350 ⫾ 52 0.75 ⫾ 0.25 89.4 ⫾ 8.0 60 ⫾ 6 7.23 ⫾ 0.06 62 ⫾ 16 14 ⫾ 4
NS NS ⬍0.01 NS NS NS ⬍0.05 NS NS ⬍0.05 ⬍0.05 ⬍0.01 ⬍0.01 ⬍0.01 ⬍0.01
Values are means ⫾ SD. PA, pulmonary artery; CVP, central venous pressure; cpm, cycles per minute; FIO2, inspired O2 fraction; SaO2, arterial O2 saturation; PaCO2, arterial PCO2; pHa, arterial pH; Cdyn, dynamic respiratory system compliance; NS, nonsignificant.
63
Räsänen J et al.
Table 2. The effects of oleic acid-induced lung injury on average acoustic power over the measurement range of 150⫺5,000 Hz, broken down by side of recording and the position of the animal during the experiment Average Power
Total Left Right Prone Supine
Baseline
Injury
P Value
⫺73.9 ⫾ 1.8 ⫺73.2 ⫾ 2.3 ⫺73.9 ⫾ 2.3 ⫺74.6 ⫾ 1.6 ⫺73.3 ⫾ 1.9
⫺71.9 ⫾ 1.9 ⫺72.6 ⫾ 1.4 ⫺72.3 ⫾ 1.9 ⫺72.0 ⫾ 2.8 ⫺71.8 ⫾ 1.0
⬍0.01 NS NS NS NS
Values are means ⫾ SD in dB.
In all cases, examination of the lungs at the end of the study revealed consolidation of the dependent part of both lungs over one-fifth to one-third of their vertical height. This finding was consistent, and its anatomical distribution reflected the randomly assigned position, prone or supine, in which the animal finished the experiment. Sound measurements. Oleic acid lung injury effected a statistically significant increase in the average spectral power of breath sounds when calculated over the entire measurement band across all six sensors (Table 2; P ⬍ 0.01). Averaging the power from the three sensors overlying each lung or over the animals studied prone or supine did not reveal statistically significant differences, or a statistically significant interaction with the effect of the injury itself. Hence, in subsequent analyses of the effect of gravity and different frequency bands, data could be averaged across both lungs and all animals. Breath sound power spectra recorded during the development of lung injury over the right dependent lung region in a representative animal are shown in Fig. 3. When the change in sound power was plotted over the studied frequency range, we noted three frequency bands in which the effect of injury appeared to be distinctly different (Fig. 4). The increase was largest within 150 – 800 Hz, moderate within 800 –3,000 Hz, and negligible at frequencies exceeding 3,000 Hz. This frequency dependency of the power change was statistically significant (P ⬍ 0.001). We also found that the frequency distribution of this power change was different for the three lung regions (Figs. 4 and 5). The increase -40 -50
Power (dB)
The results are presented as means ⫾ SD. The Wilcoxon signedrank test or Friedman’s nonparametric repeated-measures analysis of variance, where appropriate, were used to test the effects of lung injury on the measured variables. If the analysis of variance showed statistical significance, relevant individual differences between groups were tested with Dunn’s multiple-comparison test. The strength of association between changes in spectral power and physiological variables reflecting lung injury was tested with linear regression analysis performed on population averages. Results were considered statistically significant if the probability of type ␣ error was ⬍5%.
•
-60 -70 -80 -90
0
1000
2000
3000
4000
5000
Frequency (Hz) Fig. 3. Power spectra of inspiratory breath sounds recorded over the dependent portion of the right lung in one anesthetized, mechanically ventilated pig during development of oleic acid-induced lung injury. Dashed black line represents sound power distribution during baseline conditions, and solid black line during fully developed injury. The thin shaded lines represent intermediate recordings as the injury develops.
J Appl Physiol • doi:10.1152/japplphysiol.00651.2013 • www.jappl.org Downloaded from www.physiology.org/journal/jappl (109.236.055.194) on February 13, 2019.
64
Breath Sound Spectra in Experimental Lung Injury
8
8
Power (dB)
10
Power (dB)
10
6 4 2 0
•
Räsänen J et al. Supine Anterior Supine Posterior Prone Anterior
6
Prone Posterior
4 2 0
1000
-2
2000
3000
4000
5000 -2
Frequency (Hz)
1000
2000
3000
4000
5000
Frequency (Hz)
Fig. 4. Changes (⌬) in breath sound power at different frequencies resulting from oleic acid-induced lung injury recorded over nondependent (dotted line), midlung (dashed line), and dependent (solid line) locations in nine anesthetized, mechanically ventilated pigs. The response of spectral power change to injury is distinctly different in three frequency regions with boundaries at 800 and 3,000 Hz.
Fig. 6. Changes in breath sound power at different frequencies resulting from oleic acid-induced lung injury recorded over anterior and posterior chest regions in 5 supine and 4 prone, mechanically ventilated pigs. A similar change, independent of sensor location and position, is seen at a frequency band of 150 – 800 Hz. At higher frequencies, a change occurs primarily in the dependent parts of the lung, regardless of position.
was similarly large in all regions at low frequencies from 150 to 800 Hz. In the nondependent and midlung regions, the change in power was statistically significantly smaller in the 800- to 3,000-Hz (P ⬍ 0.05) and 3,000- to 5,000-Hz (P ⬍ 0.001) bands than it was at lower frequencies. In the dependent lung regions, the change was equally large in the 150- to 800-Hz and the 800- to 3,000-Hz bands, and only decreased at frequencies over 3,000 Hz (P ⬍ 0.05). Separate plots of the change in sound power in the prone and supine positions (Fig. 6) illustrate further that the increase observed in the 150- to 800-Hz range occurred in all lung areas, whereas the increase in the 800- to 3,000-Hz band developed always in the dependent areas, anterior or posterior, depending on the subject’s position prone or supine, respectively. In eight of the nine animals, we were able to compare the changes in spectral power with concurrent changes in oxygenation and lung mechanics as the lung injury was developing. The linear correlation coefficients between these variables and the three gravity-related lung regions are shown in Table 3 for two frequency bands (150 – 800 Hz, 800 –3,000 Hz). The correlation coefficients were highest for the spectral power recorded over the nondependent lung regions at the low-frequency range, and somewhat higher
for oxygenation than for lung mechanics. The time course of spectral power changes, oxygenation, and lung mechanics showed similar progression in each of the variables, starting from the first measurement 5 min after the injection of oleic acid (Fig. 7).
10
Power (dB)
8 6
*
4
*
*
2 0
DISCUSSION
Although the frequency spectra of breath sounds during spontaneous breathing in humans with normal lungs and chronic pulmonary disease have been well characterized (2), little information is available about spectral changes during mechanical ventilation and acute lung injury. Ploysongsang et al. (5) studied inspiratory frequency spectra from dogs during development of hydrostatic pulmonary edema and reported increases in mean, median, and mode frequencies that preceded the development of adventitious sounds. Vena et al. (11) recorded inspiratory breath sound spectra from healthy pigs during application of progressively higher levels of positive end-expiratory pressure and reported an attenuation and downward shifting of sound spectra that correlated with end-expiratory lung volume. In a subsequent study, Vena et al. (12) effected unilateral lung injury in pigs and recorded the appearance of pathological spectral components in a frequency range above 500 Hz in the injured lung. These changes were attributed to the appearance of crackles, they correlated with de-aeration in lung tissue, and were reversible with the application of positive end-expiraTable 3. Linear correlation coefficients (r2) between acoustic power and measures of oxygenation (VQI) and lung mechanics (Cdyn) during development of oleic acid-induced lung injury in eight pigs Frequency/Location
150-800
800-3000
3000-5000
Frequency (Hz) Fig. 5. Average differences in breath sound spectral power in three gravitational locations and three frequency bands in nine pigs between baseline conditions and oleic acid-induced lung injury. Values are means ⫾ SD. Open bars, nondependent lung; shaded bars, midlung; solid bars, dependent lung. *P ⬍ 0.05, †P ⬍ 0.001 compared with 150 – 800 Hz within the same lung region.
150–800 Hz Nondependent Midlung Dependent 800–3,000 Hz Nondependent Midlung Dependent
VQI, %
Cdyn, ml/cmH2O
0.86 0.84 0.71
0.76 0.75 0.60
0.18 0.22 0.36
0.53 0.29 0.36
VQI, ventilation-perfusion index.
J Appl Physiol • doi:10.1152/japplphysiol.00651.2013 • www.jappl.org Downloaded from www.physiology.org/journal/jappl (109.236.055.194) on February 13, 2019.
Breath Sound Spectra in Experimental Lung Injury
6
20 4
15 10
2
Power (dB)
VQI (%) Cdyn (ml/cmH20)
25
5 0
0
5 10 15 20 25 30 35 40 45 50
0
Time from injection (min) Fig. 7. Time course of changes in spectral power over nondependent lung regions (shaded bars), oxygenation [ventilation-perfusion index (VQI), solid bars], and lung mechanics [dynamic respiratory system compliance (Cdyn), open bars] during development of oleic acid-induced lung injury in eight pigs. Time zero represents baseline. Breath sound power is normalized to baseline value.
tory pressure. In both of these studies, the entire inspiratory phase of the breath was included in the spectral analysis, and no attempt was made to separate out breath sounds generated by turbulent gas flow and adventitious sounds. We designed the present study to determine whether the spectral characteristics of the breath sounds themselves, isolated from adventitious sounds, would provide an indication of injury. Such an analysis was feasible because adventitious sounds appear late in inspiration during oleic acid-induced lung injury, leaving a clean inspiratory spectral sample earlier in the inspiratory phase (Fig. 1). Monitoring of sound spectra might then provide a noninvasive means of detecting, localizing, and monitoring lung injury with an array of acoustic sensors. The results of our study revealed that oleic acid-induced lung injury induces a readily detectable increase in spectral power throughout the lung, and that this effect was commensurate with conventional measures of lung injury. Breath sounds are generated in the airways by turbulent flow, and their intensity is proportional to the gas flow rate in the airways (1). The sound waves then must propagate through the lung parenchyma, crossing the pleural space and chest wall to allow detection by a sensor on the outside surface. An increase in the power of the recorded signal implies that its original intensity has changed, the characteristics of its propagation through the respiratory system have been altered, or a new, additional signal has been added. All of these mechanisms can conceivably operate in injured lungs. Oleic acid-induced experimental lung injury disrupts the alveolar capillary membrane and causes volume loss in the injured lung (3, 10). The degree of volume loss is modified by gravity, such that the dependent lung is atelectatic and receives no gas flow, while the nondependent parts have relatively normal patency; the intermediate areas constitute a transitional spectrum of abnormalities between the two extremes (8, 9, 12). If under these circumstances the flow rate of tidal ventilation is maintained unchanged, as it was in the present study, then the gas flow rate in the airways that still conduct ventilation will become higher. Breath sounds generated in these airways will be intensified due to the
•
Räsänen J et al.
65
higher Reynolds numbers. We speculate that this is the primary mechanism for the increased acoustic power observed in the nondependent and perhaps in the midlung areas in this study. If redistribution of inspiratory gas flow were the only mechanism affecting acoustic power, then a decrease in spectral power would be expected over the dependent areas, as inspiratory flow is diverted away from them. This we did not observe; on the contrary, spectral power increase appeared to be more sustained and spread from the 150- to 800-Hz frequency range into the 800- to 3,000-Hz band in those areas. Since adventitious sounds from the spectral samples were meticulously excluded, there must have been an enhancement of sound propagation in the dependent areas to explain the increased spectral power, despite reduced air entry. We have previously shown that sound transmission through the lung parenchyma is indeed enhanced by injury in oleic acid-induced pulmonary edema in proportion to the gravity-associated involvement of any given lung region (9). This effect is likely secondary to the reduction of acoustic impedance at the lung-chest wall interface, as lung parenchyma loses gas volume. Even though the studies proving this were conducted using an external signal introduced into the airway, the same phenomenon would be expected to facilitate the conduction of natural breath sounds as well. The increase in spectral power, particularly in the nondependent areas of the lung, correlated well with conventional measures of oxygenation and lung mechanics. This is not surprising, as changes in both venous admixture and respiratory system compliance during lung injury are secondary to the development of lung regions with zero or low ventilation-to-perfusion ratios. The fact that changes in these three variables originate from the same underlying pathology indicates that their response to the progression or regression of injury will be similar, which indeed was evident in our study. Thus the monitoring of spectral changes would not be expected to provide a time advantage over monitoring oxygenation or lung mechanics. Compared with other acoustic characteristics, such as adventitious sounds, however, the early inspiratory spectral changes do occur earlier in our and others’ experience (5, 11). Our results show that early inspiratory spectral power of breath sounds increases during experimental oleic acidinduced lung injury at a rate commensurate with deterioration of oxygenation and respiratory system mechanics. Since these changes can be detected in all areas of the lung, we postulate a dual mechanism of inspiratory flow redistribution and enhanced propagation of acoustic energy to explain our observations. GRANTS This work was supported by a CR 20 grant from the Mayo Foundation. DISCLOSURES No conflicts of interest, financial or otherwise, are declared by the author(s). AUTHOR CONTRIBUTIONS Author contributions: J.R. and N.G. conception and design of research; J.R. and N.G. performed experiments; J.R., M.E.N., and N.G. analyzed data; J.R., M.E.N., and N.G. interpreted results of experiments; J.R. prepared figures; J.R. drafted
J Appl Physiol • doi:10.1152/japplphysiol.00651.2013 • www.jappl.org Downloaded from www.physiology.org/journal/jappl (109.236.055.194) on February 13, 2019.
66
Breath Sound Spectra in Experimental Lung Injury
manuscript; J.R., M.E.N., and N.G. edited and revised manuscript; J.R., M.E.N., and N.G. approved final version of manuscript. REFERENCES 1. Gavriely N, Nissan M, Rubin AHE, Cugell DW. Spectral characteristics of chest-wall breath sounds in normal subjects. Thorax 50: 1292–1300, 1995. 2. Gavriely N, Cugell DW. Airflow, amplitude and spectral content of normal breath sounds. J Appl Physiol 80: 5–13, 1996. 3. Neumann P, Hedenstierna G. Ventilation-perfusion distributions in different porcine lung injury models. Acta Anaesthesiol Scand 45: 78 –86, 2001. 4. Pelosi P, D’Andrea L, Vitale G, Pesenti A, Gattinoni L. Vertical gradient of regional lung inflation in adult respiratory distress syndrome. Am J Respir Crit Care Med 149: 8 –13, 1994. 5. Ploysongsang Y, Michel RP, Rossi A, Zocchi L, Milic-Emili J, Staub NC. Early detection of pulmonary congestion and edema in dogs by using lung sounds. J Appl Physiol 66: 2061–2070, 1989. 6. Räsänen J, Downs JB, Hodges MR. Continuous monitoring of gas exchange and oxygen use with dual oximetry. J Clin Anesth 1: 3–8, 1988.
•
Räsänen J et al.
7. Räsänen J, Downs JB, Malec DJ, DeHaven B, Garner WL. Real-time continuous estimation of gas exchange by dual oximetry. Intensive Care Med 14: 118 –122, 1988. 8. Räsänen J, Gavriely N. Detection of porcine oleic acid-induced acute lung injury using pulmonary acoustics. J Appl Physiol 93: 51–57, 2002. 9. Räsänen J, Gavriely N. Response of acoustic transmission to positive airway pressure therapy in experimental lung injury. Intensive Care Med 31: 1434 –1441, 2005. 10. Rosenthal C, Caronia C, Quinn C, Lugo N, Sagy M. A comparison among animal models of acute lung injury. Crit Care Med 26: 912–916, 1998. 11. Vena A, Perchiazzi G, Giuliani R, Fiore T, Hedenstierna G. Acoustic effects of positive end-expiratory pressure on normal lung sounds in mechanically ventilated pigs. Clin Physiol Funct Imaging 26: 45–53, 2006. 12. Vena A, Rylander C, Perchiazzi G, Giuliani R, Hedenstierna G. Lung sound analysis correlates to injury and recruitment as identified by computed tomography: an experimental study. Intensive Care Med 37: 1378 – 1383, 2011.
J Appl Physiol • doi:10.1152/japplphysiol.00651.2013 • www.jappl.org Downloaded from www.physiology.org/journal/jappl (109.236.055.194) on February 13, 2019.
