OBJECTIVE: To determine whether breath sound distribution maps can differentiate between patients with pneumonia or pleural effusion versus healthy controls. METHODS: We recorded breath sounds from 20 patients conventionally diagnosed as having pleural effusion, 20 patients conventionally diagnosed as having pneumonia, and 60 healthy controls, of whom 20 served as a learning sample. All subjects were examined with a computer-based multi-sensor breath sound mapping device that records, analyzes, and displays a dynamic map of breath sound distribution. The physicians who interpreted the breath sound images were first trained in identifying common characteristics of the images from the learning sample of normals. Then the images from the 40 patients and the 40 controls were interpreted as either normal or abnormal. RESULTS: In the normal images, the left and right lung images developed synchronously and had similar size, shape, and intensity. The sensitivity and specificity of blinded differentiation between normal and abnormal images when the physician interpreter did not know the patient's workup were 82.5% and 80%, respectively. The sensitivity and specificity of blinded detection of normal and abnormal images when the interpreter did know the patient' workup were 90% and 88%, respectively. CONCLUSIONS: Computerized dynamic imaging of breath sounds is a sensitive and specific tool for distinguishing pneumonia or pleural effusion from normal lungs. The role of computerized breath sound analysis for diagnosis and monitoring of lung diseases needs further evaluation.
OBJECTIVE: To determine whether breath sound distribution maps can differentiate between patients with pneumonia or pleural effusion versus healthy controls. METHODS: We recorded breath sounds from 20 patients conventionally diagnosed as having pleural effusion, 20 patients conventionally diagnosed as having pneumonia, and 60 healthy controls, of whom 20 served as a learning sample. All subjects were examined with a computer-based multi-sensor breath sound mapping device that records, analyzes, and displays a dynamic map of breath sound distribution. The physicians who interpreted the breath sound images were first trained in identifying common characteristics of the images from the learning sample of normals. Then the images from the 40 patients and the 40 controls were interpreted as either normal or abnormal. RESULTS: In the normal images, the left and right lung images developed synchronously and had similar size, shape, and intensity. The sensitivity and specificity of blinded differentiation between normal and abnormal images when the physician interpreter did not know the patient's workup were 82.5% and 80%, respectively. The sensitivity and specificity of blinded detection of normal and abnormal images when the interpreter did know the patient' workup were 90% and 88%, respectively. CONCLUSIONS: Computerized dynamic imaging of breath sounds is a sensitive and specific tool for distinguishing pneumonia or pleural effusion from normal lungs. The role of computerized breath sound analysis for diagnosis and monitoring of lung diseases needs further evaluation.