JOURNAL HOME CME HOME THIS MONTH PAST ISSUES ETOC COLLECTIONS AUTHORS REVIEWERS EDITORIAL BOARD FEEDBACK RSS HELP
		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (1) Citing Articles via Google Scholar Google Scholar Articles by Tejman-Yarden, S. Articles by Gurman, G. M. Search for Related Content PubMed PubMed Citation Articles by Tejman-Yarden, S. Articles by Gurman, G. M. Related Collections Monitoring (Non-cardiac) Technology

---

TECHNOLOGY, COMPUTING, AND SIMULATION

Acoustic Monitoring of Double-Lumen Ventilated Lungs for the Detection of Selective Unilateral Lung Ventilation

Shai Tejman-Yarden, MD^*, Dror Lederman, MSc, Israel Eilig, MD, Alexander Zlotnik, MD, Nathan Weksler, MD, Arnon Cohen, PhD (RIP), and Gabriel M. Gurman, MD

From the *Division Of Pediatrics, Soroka University Medical Center, Beer Sheva, Israel; Department of Electrical and Computer Engineering, Ben Gurion University of the Negev, Beer Sheva, Israel; and Division of Anesthesiology and Critical Care, Soroka Medical Center, Beer Sheva, Israel.

Address correspondence and reprint requests to Shai Tejman-Yarden, MD, Division of Pediatrics, Soroka Medical Center, Faculty of Health Sciences, Ben Gurion University of the Negev, Beer Sheva 84101, Israel. Address e-mail to teyman{at}bgumail.bgu.ac.il or tegmanya{at}inter.net.il.

	Abstract

Top Abstract Introduction METHODS RESULTS DISCUSSION REFERENCES

 
One-lung intubation (OLI) is among the most common complications of endotracheal intubation. None of the monitoring tools now available has proved effective for its early detection. In this study we investigated the efficacy of acoustic analysis for the detection of OLI. We collected lung sounds from 11 patients undergoing thoracic surgery requiring the placement of a double-lumen tube. Recordings of separate lung ventilation were performed after induction and confirmation of adequate tube positioning, before surgery. Samples of lung sounds were collected by three piezoelectric microphones, one on each side of the chest and one on the right forearm, for background noise sampling. The samples were filtered, the signals’ energy envelopes were calculated, and segmentation to breath and rest periods was performed. Each respiration was classified into one of the three categories: bilateral ventilation, selective right-lung ventilation, or selective left-lung ventilation, on the basis of the ratio between the energy signals of each lung. OLI was accurately identified in 10 of the 11 patients during right OLI and in all 11 patients during left OLI. This study suggests that acoustic monitoring is effective for the detection of selective lung ventilation and may be useful for early diagnosis of OLI.

	Introduction

Top Abstract Introduction METHODS RESULTS DISCUSSION REFERENCES

 
Inadvertent endobronchial intubation or one-lung intubation (OLI) was reported as the most common complication of endotracheal intubation by the Australian Incident Monitoring Study Analysis (1). It accounted for 79 of the 189 reported incidents. OLI is particularly common during pediatric anesthesia and laparoscopic surgery (2,3) and may lead to serious complications, such as atelectasis of the nonventilated lung, pneumothorax, hypoxemia, cardiac arrhythmias, and hypotension (4). None of the techniques used to monitor intubated patients has proved reliable for the early detection of OLI before the appearance of clinical signs.

Auscultation of breath sounds, used in daily practice to confirm the correct positioning of the endotracheal tube in the operating room or in the intensive care unit, was found to be inaccurate for the detection of OLI (5–7), with an unacceptably high margin of error that can reach up to 60% (4). Pulse oxymetry for oxygen saturation monitoring is considered a good method for OLI detection (8,9). The Australian Incident Monitoring Analysis reported that pulse oxymetry was used to detect 87% of the OLI incidents. But oxymetry results have a latency of 2–5 min, becoming apparent only after the patient becomes hypoxemic, without providing any indication as to the reason of the hypoxemia (1). Capnography was suggested in the past, but has proved to be a poor method to identify OLI (10,11).

New techniques for ventilation monitoring are being developed and studied (12,13). The Video Stethoscope was reported once as a method for assuring continuous bilateral lung ventilation (12) and online spirometry has been studied and reported to have several limitations (14). Acoustic reflectometry, a method based on an area– distance profile used to distinguish between tracheal and esophageal intubations (15), was found to be useful in the detection of OLI in a patient undergoing laparoscopic surgery (16), but no further validation of this method has been provided.

A preliminary study by Sod-Moriah et al. (17,18) demonstrated that noninvasive monitoring of lung sounds by placement of a microphone on each side of the chest was useful for the monitoring of differential lung ventilation in dogs, detecting 89% of the cases with selective right (Rt) OLI and 90% of the cases with selective left (Lt) OLI (19,20). These encouraging results lead us to try this monitoring technique on human patients. To sample selective OLI, we sampled patients intubated for general anesthesia with a double- lumen tube. This ventilation technique enabled us to obtain lung sounds during only Rt, only Lt, and bilateral-lung ventilation by clamping one side of the tube and ventilating the other.

The aim of the present study was to validate the accuracy of an acoustic sensor device for the detection of selective OLI in human patients ventilated with a double-lumen endotracheal tube.

	METHODS

Top Abstract Introduction METHODS RESULTS DISCUSSION REFERENCES

 
This pilot study was approved by the ethics committee of the Soroka Medical Center, Ben-Gurion University of the Negev, Beer-Sheva, Israel.

Patients
Eleven adult surgical patients, ASA I–II, were included in this study and informed consent was received from all. The study patients were all scheduled for a surgical lung procedure requiring the insertion of a double-lumen endotracheal tube. The tube used in this trial was a Rüsch double-lumen tube, with the endobronchial tube positioned in the Lt bronchus.

The Sampling System
The sampling system consisted of three Tommyscope acoustic sensors (Innovative Medical Equipment, KOL Medical, Ramat-Gan, Israel), an amplifier, an antialiasing filter, and a laptop computer which was equipped with an A/D converter. For safety reasons, rechargeable batteries were used for the amplifier and the computer. The Tommyscope sensors are piezoelectric microphones with an internal low pass filter which suits the spectral range of the breathing sound signal. The amplifier has a controlled gain for each channel and the laptop computer is equipped with a Keithley A/D converter with a 12-bit resolution.

Technique
Before anesthesia three Tommyscope acoustic sensors were attached to each patient. Two microphones were placed on the chest, in the mid-clavicular line (one on each side of the chest) and a third microphone was placed on the right forearm to sample the background noise. A double-lumen tube was placed in position after induction of anesthesia. The correct positioning of the endobronchial tube in the Lt main bronchus was confirmed with a fiberoptic bronchoscope. After anchoring the tube to the patient and inflating both Lt bronchial and mainstem balloons, sets of lung sounds were obtained sequentially from each patient in the following order: bilateral-lung ventilation, Lt lung ventilation only, and Rt lung ventilation only. Unilateral ventilation was achieved by clamping one side of the double-lumen tube and ventilating the contralateral side (OLI). No clamping was performed during bilateral-lung ventilation. Although all the patients were recorded in the same order, (Lt, Rt, and bilateral), for each patient a different protocol was given, in which a different number of respirations from each side was designed. The recordings were completed before the commencement of surgery and did not interfere with the surgical procedure.

Signal Preprocessing
The recorded sound signals were preamplified through a 1.6 kHz antialiasing filter and sampled at 4.0 kHz using 12-bit A/D. Processing of the sampled signal was performed using the MATLAB software (Mathworks, Natick, MA). The signals were first down sampled to 1.0 kHz and then filtered using a 14 order IIR Chebyshev band pass filter in the range of 125–200 Hz to reduce the background noise and the interference of heart sounds. The filtered signals were divided into 20-ms segments and for each segment the energy was calculated to obtain the signals’ energy envelope. On the basis of the energy envelopes, segmentation of the lung sounds to breath and rest periods was performed. The signal of the background noise sampled by the microphone placed on the right forearm was not used in the current implementation because of the relatively high signal-to-noise ratio of the recorded lung sound signals and adequate noise filtration by the Tommyscope sensors and the band pass filter.

Signal Processing and Breath Classification
For identification of the ventilation status of each patient, we used an algorithm which is based on the assumption that the mutual interference between the two lungs is relatively small, and that each microphone represents the sound of its ipsilateral lung. This algorithm classifies the lung sounds into three classes: tracheal or bilateral ventilation (Tr), selective Lt lung ventilation (Lt OLI), or selective Rt lung ventilation (Rt OLI). The classification is done by calculating the ratio of the energy envelopes of every breath, E_L/E_R (the Lt energy envelope divided by the Rt energy envelope). A ratio above the higher threshold indicates Lt OLI, a ratio below the lower threshold indicates Rt OLI, and all the energy ratios between the two thresholds indicate bilateral ventilation. It should be noted that the comparison of energy envelopes is a strategy often used in pattern recognition problems (21). Since the database was small and each sample had a significant impact on the algorithm’s basic threshold template, to increase the resolution and the reliability of this trial we used the leave-one-out method. Using this method, for each patient the energy ratios of the sampled lung sounds were compared to a threshold template calculated from the samples of the other 10 patients, which served as a reference database. This process was repeated 11 times, for each patient separately so that the tested data would not have any influence on the algorithm. The basic threshold template reported here is the mean of all experiments from which the energy ratios were set. This template reveals that the higher and lower thresholds are 11 and 0.3 respectively—an asymmetry between the Rt and the Lt lung sounds which has been described by Cohen and co-workers (17–22).

It should be mentioned that signal processing was performed by an electrical engineer who was not present in the operating room during anesthesia and surgery, and was blinded to the true ventilation protocol of each data sample. The results of the classification calculated by the computer were compared with the true ventilation status protocol only at the end of the trial.

	RESULTS

Top Abstract Introduction METHODS RESULTS DISCUSSION REFERENCES

 
After induction of anesthesia and fiberoptic positioning of the double-lumen tube, each patient was sampled during Tr-lung ventilation, Lt-lung ventilation, and Rt-lung ventilation. All the samples were filtered and processed. An example of the filtration needed for the sampled sounds is shown in Figure 1. Boxes a and c present the sounds sampled from the Rt lung; a before and c after noise filtration. Boxes b and d represent the sounds sampled from the Lt lung; b before and d after noise filtration. The filtered recordings from each side of the chest show how well the recorded samples demonstrate the patient’s ventilation status. As seen in box c, the filtered recording of the Rt lung begins with sampled breath sounds (bilateral ventilation). After 23 s the lung sounds are diminished during Lt OLI and after 45 s loud lung breathing sounds are recorded during Rt OLI. In box d, the filtered recording of the Lt lung, the sample begins with sampled breathing sounds (bilateral ventilation). After 23 s during Lt OLI the lung sounds are louder and after 45 s the lung sounds are diminished during Rt OLI.

View larger version (42K): [in this window] [in a new window]   Figure 1. Patterns of sampled lung sounds recorded from the patients: box a is an unfiltered lung sound sample from the right chest, box c is the same sample after filtration. Box b is a recording of the same trial from the left chest and box d is the same sample after filtration. For the first 23 s both lungs were ventilated, after 23 s the right tube was clamped and only the left lung was ventilated, and after 45 s the right tube was unclamped and the left side tube was clamped so that only the right lung was ventilated.

An example of the Matlab calculations done on the signals in Figure 1 is shown in Figure 2. In this figure boxes c and d show the calculated energy envelopes. The lung sound classification is printed between the boxes. The computerized recognition algorithm successfully recognized the different ventilation conditions and classified each respiration correctly.

View larger version (35K): [in this window] [in a new window]   Figure 2. Signal processing of the sample in Figure 1. In boxes a and b the samples are shown in greater detail after segmentation. In boxes c and d the energy envelope is shown for each respiration. The units of the horizontal axis in boxes c and d are frames. Each frame measures 20 ms, for which the energy envelope is calculated. Each respiration is classified into Tr (tracheal ventilation), Rt OLI (right one-lung intubation) and Lt OLI (left OLI). The classification for each respiration appears between the boxes of the right and left samples.

Lt OLI was correctly identified in all 11 patients, but Rt OLI was correctly diagnosed in only 10 of 11 patients. In the remaining patient the classification algorithm indicated a bilateral-lung ventilation status during Rt OLI. This false negative result is shown in Figure 3. In this case, the first breath is bilateral ventilation. After this, the Rt tube was clamped and for 35 s only the Lt lung was ventilated, after which the Rt tube was unclamped and the Lt tube was clamped so that only the Rt lung was ventilated. In boxes a and b the filtered sound samples are presented and in boxes c and d the calculated energy envelopes of the sampled sounds are presented. Between these boxes the computerized recognition of the ventilation status is designated. In this trial, the classification algorithm incorrectly indicated bilateral lung ventilation during Rt OLI, 40 s after the trial began.

View larger version (41K): [in this window] [in a new window]   Figure 3. Recordings of lung sounds and signal processing. In this trial after 40 s, during right (Rt) one-lung intubation the algorithm incorrectly classified the respiration as tracheal (Tr).

Table 1 shows a confusion matrix of the results of the classification algorithm compared to the experimental protocol. The overall sensitivity of the system was 32 correct readings of 33 samples, or 97%, which represents the positive predictive value for this algorithm. There was 100% recognition of Lt OLI (11 correct readings of 11), for a sensitivity of 100%, whereas, the recognition rate for Rt OLI was 92% (10 correct readings of 11 samples), for a sensitivity of 90%.

	DISCUSSION

Top Abstract Introduction METHODS RESULTS DISCUSSION REFERENCES

 
Intubation of either of the main bronchi is a complication of endotracheal anesthesia (1). Auscultation, which relies upon the same signals processed in this work, has been shown repeatedly to be an unreliable technique for detection of proper bilateral ventilation (4–7). Despite this, Sod Moriah et al. (18) showed an 89% recognition rate of Rt OLI and 90% recognition rate of Lt OLI using an algorithm based on lung sound analysis in a preliminary study of anesthetized dogs. On the basis of their findings, we used an acoustic technique to sample lung sounds from human patients undergoing surgical procedures which required the insertion of a double-lumen tube, and which enabled us to sample each patient for Rt OLI, Lt OLI, and bilateral-lung ventilation.

The analysis of the lung sounds, using an algorithm which assumes that each microphone samples its ipsilateral lung, yielded 100% recognition of Lt OLI with a sensitivity of 100% and 92% recognition of Rt OLI with a sensitivity of 90%. These results correspond to those reported by Sugiyama and Yokoyama (7) who studied auscultation sounds during Rt endobronchial intubation and assumed that gas passing through the Murphy’s eye or through a narrow space between the endotracheal tube and the bronchial wall during selective endobronchial intubation might generate contralateral sound signals. In our study we used a double-lumen tube with inflated balloons so that a significant amount of air leak would not account for our findings.

Careful analysis of the sound samples collected in this study shows that each chest microphone actually sampled sounds from both lungs, but with differential ipsilateral and contralateral contributions. Moreover, we found that each lung does not produce a uniform sound, but a series of various sounds at different times during each respiration. This brings us to the conclusion that this simple algorithm, which assumes that each microphone represents its ipsilateral lung, does not adequately reflect this multiple input, multiple output system. This is probably why repeated trials to monitor lung ventilation by auscultation to breath sounds, with a stethoscope or a set of microphones, failed to detect OLI. Altering this algorithm by narrowing the ratio margins may raise the sensitivity for Rt and Lt OLI; however, it may decrease the specificity and false alarms will appear. A more comprehensive multisensor approach using a proper processing algorithm, which considers the transmission of sound from each lung to the other, may increase the accuracy of tube positioning detection during artificial ventilation.

In conclusion, this preliminary study suggests that an acoustic monitoring technique which is noninvasive may be useful for confirmation of the endotracheal tube’s position and for early detection of inadvertent OLI. Further studies are now underway to validate this hypothesis, to develop a better algorithm for the separation of lung sounds and to recognize improper ventilation. A new device that may alert the anesthesiologist of ventilation faults before the development of any life threatening symptoms is a goal which can be achieved.

	ACKNOWLEDGMENTS

 
While working on this project, Professor Arnon Cohen, head of the signal processing laboratory in Ben-Gurion University had died. Prof. Cohen was known world-wide as a researcher who wrote numerous articles, chapters and books, taught hundreds of students, and led a vibrant laboratory. By completing this work we continue his legacy of honest and fair research.

	Footnotes

 
Accepted for publication August 1, 2006.

Deceased. See Acknowledgments.

	REFERENCES

Top Abstract Introduction METHODS RESULTS DISCUSSION REFERENCES

 

1. Szekely SM, Webb RK, Williamson JA, Russell WJ. The Australian Incident Monitoring Study. Problems related to the endotracheal tube: an analysis of 2000 incident reports. Anesth Intensive Care 1993;21:611–6.[ISI][Medline]
2. Rivera R, Tibballs J. Complications of endotracheal intubation and mechanical ventilation in infants and children. Crit Care Med 1992;20:193–9.[ISI][Medline]
3. Joshi GP. Complications of laparoscopy. Anesthesiol Clin North Am 2001;19:89–105.[Medline]
4. Brunel W, Coleman DL, Schwartz DE, et al. Assessment of routine chest roentgenograms and the physical examination to confirm endotracheal tube position. Chest 1989;96:1043–5.[ISI][Medline]
5. Schwartz DE, Lieberman JA, Cohen NH. Women are at greater risk than men for malpositioning of the endotracheal tube after emergent intubation. Crit Care Med 1994;22:1127–31.[ISI][Medline]
6. Klepper ID, Webb RK, Van der Walt JH, et al. The Australian Incident Monitoring Study. The stethoscope: applications and limitations—an analysis of 2000 incident reports. Anesth Intensive Care 1993;21:575–8.[ISI][Medline]
7. Sugiyama K, Yokoyama K. Reliability of auscultation of bilateral breath sounds in confirming endotracheal tube position. Anesthesiology 1995;83:1373.[ISI][Medline]
8. Webb RK, van der Walt JH, Runciman WB, et al. The Australian Incident Monitoring Study. Which monitor? An analysis of 2000 incident reports. Anesth Intensive Care 1993;21:529–42.[ISI][Medline]
9. Webster TA. Now that we have pulse oximeters and capnographs, we don’t need precordial and esophageal stethoscopes. J Clin Monit 1987;3:191–2.[ISI][Medline]
10. Heneghan CP, Scallan MJ, Branthwaite MA. End-tidal carbon dioxide during thoracotomy. Its relation to blood level in adults and children. Anaesthesia 1981;36:1017–21.[ISI][Medline]
11. Johnson DH, Chang PC, Hurst TS, et al. Changes in PETCO₂ and pulmonary blood flow after bronchial occlusion in dogs. Can J Anaesth 1992;39:184–91.[Abstract/Free Full Text]
12. Huang KC, Kraman SS, Wright BD. Video stethoscope—a simple method for assuring continuous bilateral lung ventilation during anesthesia. Anesth Analg 1983;62:586–9.[Abstract/Free Full Text]
13. Manecke GR Jr, Hartman AR, Poppers PJ. Computer-assisted auscultation of a bronchopleurocutaneous fistula during general anesthesia. Anesth Analg 1996;83:880–2.[ISI][Medline]
14. Bardoczky GI, Levarlet M, Englelman E, deFrancquen P. Continuous spirometry for detection of double-lumen endobronchial tube displacement. Br J Anaesth 1993;70:499–502.[Abstract/Free Full Text]
15. Raphael DT, Lee H. Acoustic reflectometry detection of an endobronchial intubation in a patient with equal breath sounds. J Clin Anesth 2003;15:41–7.[ISI][Medline]
16. Raphael DT, Benbassat M, Arnaudov D, et al. Validation study of two-microphone acoustic reflectometry for determination of breathing tube placement in 200 adult patients. Anesthesiology 2002;97:1371–7.[ISI][Medline]
17. Sod-Moriah G. Ventilation monitoring during anesthesia and respiratory intensive care. Beer-Sheva, Israel: Department of Electrical and Computer Englineering, Ben-Gurion University of the Negev, 1995.
18. Sod-Moriah G, Cohen A, Gurman GM. Detection of one-lung intubation incidents in general anesthesia and intensive care. In: Proceedings of the 13th International Conference of Biosignalling 96, Brno, Czech Republic, 1996.
19. Cohen A, Landsberg D. Analysis and automatic classification of breath sounds. IEEE Trans Biomed Engl 1984;31:585–90.
20. Cohen A, Berstein AD. Acoustic transmission of the respiratory system using speech stimulation. IEEE Trans Biomed Engl 1991;38:126–32.
21. Cohen A. Biomedical signal processing. Boca Raton, FL: CRC Press, 1986.
22. Wodicka GR, DeFrain PD, Kraman SS. Bilateral asymmetry of respiratory acoustic transmission. Med Biol Engl Comput 1994;32:489–94.

This article has been cited by other articles:

				D. Steinmann, C. A. Stahl, J. Minner, S. Schumann, T. Loop, A. Kirschbaum, H. J. Priebe, and J. Guttmann Electrical impedance tomography to confirm correct placement of double-lumen tube: a feasibility study Br. J. Anaesth., September 1, 2008; 101(3): 411 - 418. [Abstract] [Full Text] [PDF]

				S. Tejman-Yarden, A. Zlotnik, L. Weizman, J. Tabrikian, A. Cohen, N. Weksler, and G. M. Gurman Acoustic Monitoring of Lung Sounds for the Detection of One-Lung Intubation Anesth. Analg., August 1, 2007; 105(2): 397 - 404. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (1) Citing Articles via Google Scholar Google Scholar Articles by Tejman-Yarden, S. Articles by Gurman, G. M. Search for Related Content PubMed PubMed Citation Articles by Tejman-Yarden, S. Articles by Gurman, G. M. Related Collections Monitoring (Non-cardiac) Technology

Anesthesia & Analgesia® is published for the International Anesthesia Research Society® by Lippincott Williams & Wilkins with the assistance of Stanford University Libraries' HighWire Press®. Copyright 2006 by the International Anesthesia Research Society. Online ISSN: 1526-7598   Print ISSN: 0003-2999