Analysis of Quantitative Sound Characteristics for Detection of Lung Obstruction with Secretions

• Main Authors: Chieh-Hsun Chang-Chien, 張簡介勛
• Other Authors: Chun-Ju Hou
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/83343727220876552515

碩士 === 南台科技大學 === 電機工程系 === 98 === Breathing sounds may reflect the situation of the air flow in the trachea and bronchial system. In addition, it can also identify the location of the expansion or blockage and asses the status of the lung expansion in peripheral alveoli and pleural cavity. An increase of secretions, obstruction or stenosis in respiratory tracts will generate abnormal breath sounds. During physical examination, stethoscope is the most common, inexpensive, and convenient medical equipment. However, the decline in auscultation skills of medical personnel may cause inaccurate judgment of the symptoms. The purpose of this study is to establish signal processing, analysis and recognition system of lung sounds and breathing patterns for assisting nurses in evaluating the condition of the respiratory system and determining the timing of suction. Signal processing in the system includes signal detrending for eliminating the baseline drift and adaptive filtering for removing the background interference. The signal analysis includes the frequency domain analysis, the time-frequency analysis and wavelet analysis. The statistical method, the discriminant analysis, and the artificial intelligent method, the backpropagation neural network, are used for establishing the recognition models of lung sounds and breathing patterns. The performance of two methods is tested and compared. The results show that the accuracy of the lung sound recognition is 91.6% in the discriminant analysis and 94.5% in the backpropagation neural network. The accuracy of the breathing pattern recognition is 42.8%. The study concludes that our system is feasible to discriminate six lung sounds and the performance is bad in the breathing pattern recognition. In the future work, the improvement in our system includes the algorithm of the parameter extraction and the design of the pressure sensor pad.