Acoustic monitoring of lung sounds for the detection of one lung intubation

Background: One lung intubation (OLI) is a major incident during endotracheal intubation. Monitoring methods used today have been found to be non specific and controversial regarding its early diagnosis. We have developed a new system which detects OLI based on electronic lung sounds analysis during artificial ventilation. Methods: 24 adult surgical patients schedualed for routine surgical procedures were included. Lung sound sampling was done during induction to anesthesia and tube positioning by four piezoelectric microphones attached to the patients' back. To achieve OLI, the endotracheal tube was inserted and advanced down the airway so that no left breath sounds were heard; the tube was then withdrawn stepwise until equal breath sounds could be heard. The final position of the tube was confirmed by fiberoptic bronchoscopy. The sampled lung sound were processed by an algorithm which assumes a MIMO (Multi Input Multi Output) system; in which a multi-dimensional auto-regresive (AR) model relates the input (lungs) and the output (recorded sounds). A detector for OLI was developed, based on a Generalized Likelihood Ratio Test, which was implemented under a coherent sources assumption. Results: Based on this new algorithm the probability of OLI detection is 95.2% with a probability of 4.8% false alarm. We can assume even higher values of detection, depending on the sensitivity wanted, taking in mind higher incidence of false alarms. Summary: We present a new device, already in prototype, which detects OLI in various clinical scenarios with a very high degree of accuracy.