Pneumothorax is a potentially life-threatening condition, especially in critically ill patients. Although usually diagnosed clinically, physical signs may be equivocal, thus leading to delayed diagnosis and possible adverse clinical outcomes. Chest radiography may not always establish the diagnosis, and CT scanning may be impractical given cost, availability and scheduling considerations. We hypothesized that pneumothorax can be accurately detected by computerized analysis of breath sounds.
After IRB approval and informed consent, breath sounds of 10 subjects (5 controls and 5 with pneumothorax) were recorded using an electronic stethoscope. Subjects were studied in the semi-upright position with the stethoscope placed over the mid clavicular line just inferior to the clavicle. Breath sound signals were digitized and “bandpass” filtered (300-700 Hz ) using a personal computer. The breath sound amplitude variation was calculated as the ratio between mean and minimum sound amplitude during inspiration. This ratio was compared between the control and pneumothorax groups using the Wilcoxon signed ranks test.
The inspiratory breath sound variability was smaller in the pneumothorax group, which is consistent with decreased airflow and airflow variation with pneumothorax. Using this criterion, 100% separation of the pneumothorax and control group was possible (p < 0.008).
Pneumothorax may be accurately detected by computerized analysis of breath sounds, suggesting that further research into these methods is warranted.
Simple bedside devices may use computerized breath sound analysis to rapidly alert clinicians to the occurrence of a pneumothorax, leading to more rapid diagnosis and earlier intervention.
H.A. Hassaballa, None.