USING DEEP LEARNING TO ANALYZE BREATH SOUNDS FOR RESPIRATORY ILLNESS REORGANIZATION
Asian Journal of Advances in Research,
Many common respiratory illnesses like bronchitis, asthma, and COPD lead to bronchial inflammation, but it's hard to tell the timing and severity of the blockage. We hope to create and improve a deep learning program that can constantly report the bronchi inflammation by recording their breath sounds in a non-intrusive/uncomplicated way. This research would give doctors a better understanding of what treatment to use and allow faster treatment. Our purpose is to develop a deep learning program that recognizes bronchi blockages through sound recordings at a >70% success rate. We printed various-shaped bronchi blockages for my methods and collected approximately 20 one-second audio clips with voltage at max. Then, we cut all audio clips into one-second clips in Audacity and duplicated sound records to make ~960 for each blockage.
Lastly, we trained deep learning models and tested the success rate. From our observations, the deep learning program had a 99.28% recognition rate when inputting data for eight blockages. Despite moving the recording device every five recordings, a high recognition rate was achieved, which means that the deep learning program can account for certain external factors like the location of the recording device. Human testing would require many more samples and more external factors, and our current program only works for pre-recorded sounds. However, it will be possible to implement my deep learning model into real-world applications so that in vivo breath sounds are used to approximate the size and shape of the bronchi blockages instantly.
- Deep learning
- respiratory illness
How to Cite
Rhedin S, Lindstrand A, Rotzén-Östlund M, Tolfvenstam T, Öhrmalm L, Rinder MR, Naucler P. Clinical utility of PCR for common viruses in acute respiratory illness. Pediatrics. 2014;133(3):e538-e545.
Blanc PD, Annesi-Maesano I, Balmes JR, Cummings KJ, Fishwick D, Miedinger D, Redlich CA. The occupational burden of nonmalignant respiratory diseases. An official American Thoracic Society and European Respiratory Society statement. American journal of respiratory and critical care medicine. 2019;199(11):1312-1334.
Shakhgeldyan K, Geltser B, Kurpatov I, Kriger A. Artificial Intelligence Methods in Assessing the Severity and Differential Diagnosis of Bronchoobstructive Syndrome. In 21st International Workshop on Computer Science and Information Technologies CSIT. 2019; 74-78. Atlantis Press.
Muthusamy PD, Sundaraj K, Abd Manap N. Computerized acoustical techniques for respiratory flow-sound analysis: a systematic review. Artificial Intelligence Review. 2020;53(5):3501-3574.
Walker CM, Rosado-de-Christenson ML, Martínez-Jiménez S, Kunin JR, Wible BC. Bronchial arteries: anatomy, function, hypertrophy, and anomalies. Radiographics. 2015;35(1):32-49.
Abstract View: 122 times
PDF Download: 24 times