Analysis of tracheal sounds acquired from patients with lung cancer

Abstract

In this study an attempt has been made to determine the quantitative characteristics of the sounds which are caused by the presence of a tumor in the bronchial tree. For this purpose, a special instrumentation system consisting of a PC equipped with an A/D converter, a water-sealed spirometer and amplifiers for tracheal sounds, environmental sounds and a flow potentiometer have been used. Using this system the tracheal sounds of the subject, environmental noise and flow rate of the breathing of the subject were recorded simultaneously. Although vocal cords and oral cavity produce interfering sound signals, trachea was selected as the precise location to acquire the diagnostic lung sounds. This is because the malign tumors of the lung are mostly found in the large bronchi, respiratory sounds measured at the trachea undergo very little filtering, andcharacteristics of the tracheal lung sounds do not depend on subject's morphology to a great extend.The sound signals were filtered using high pass filters having cut-off frequencies at 100 Hz and 2500 Hz, respectively. A sampling frequency of 5 Hz was selected. 1024 data points were extracted from the proper inspiration and expiration phases of each subject. Using a Hanning window and DFFT, the Welch periodogram was estimated.Because patients with lung cancer and chronic obstructive pulmonary disease (COPD) are usually heavy smokers, they have similar symptoms. Therefore, COPD patients were also included to this study. Regions of diagnostic significance in the frequency spectra of subjects studies have been identified and confirmed using the statistical t-test. The area of such a region in the frequency spectra above 633.8 Hz in the inspiration phase of patients with lung cancer and COPD was determined as a distinctive feature with a possibility of less than 18% error.