Observer disagreement for the physical findings used to diagnose pneumonia is large. We have developed a computerized multichannel lung sound analyzer that detects and provides objective quantification of sounds from 16 sites simultaneously. Our purpose was to determine whether objectively detected lung sounds in patients with pneumonia are significantly different than those in asymptomatic age matched controls.
A convenience sample of 100 patients, age 69±18 (mean±SD), in a community teaching hospital who had a clinical diagnosis of pneumonia and 100 controls, age 69±7, were examined with a 16 channel lung sound analyzer (Stethographics, Inc., Model 1602). An Acoustic Pneumonia Score (APS) was generated based on individual acoustic findings detected by the system including rates of rhonchi, automatic counts of fine and coarse crackles as well as amplitude measurements of inspiration and expiration.
Inspiratory crackles were present in 81% of these patients as compared to 28% of controls. Expiratory crackles were also more common-65% as compared to 9%. Wheezes and rhonchi were more common in the patients with pneumonia – 18% had wheezing or rhonchi in inspiration and 19% in expiration as compared to 1% and 0% respectfully in the controls. The APS averaged 13±8 in patients with pneumonia and 3±3 in controls, p<0.0001. The positive predictive power of a score higher than 6 was 0.90. Its sensitivity was 0.85 and specificity was 0.91.
Lung sound analysis can provide objective evidence supporting the diagnosis of pneumonia. The method is noninvasive and easy to perform even in severely ill patients on ventilators.
Automated lung sound anlysis offers the promise of assisting clinicians in the diagnosis of pneumonia particulary in the ICU where chest x-ray may be suboptimal (e.g. retrocardiac pneumonias and those below the diaphragmatic dome).
R. Murphy and A. Vyshedskiy have a financial interest in Stethographics.