Estimation of Inhalation Flow Parameters for Asthma Monitoring Using Acoustic Signal Processing and Machine Learning

• Autores: Zineb Jeddi, Mounir Ghogho, Adam Bohr, Johan P. Botker, Ismail Kassou
• Localización: Intelligent environments 2019: Workshop Proceedings of the 15th International Conference on Intelligent Environments / coord. por Andrés Muñoz Ortega, Sofia Ouhbi, Wolfgang Minker, Loubna Echabbi, Miguel Navarro Cía, 2019, ISBN 9781614999836, págs. 317-325
• Idioma: inglés
• Texto completo no disponible (Saber más ...)
• Resumen

  • Asthma is one of the most prevalent diseases. To control this chronic respiratory condition, inhaler devices, such as the dry powder inhaler (DPI) and the pressurized metered dose inhaler (pMDI), are prescribed. However, poor asthma management can significantly deteriorate the patients’ health and their quality of life in general. Through regular treatment, asthma patients frequently use inhaled medication in order to help improving their respiratory system. Nevertheless, a lot of patients do not follow the inhalation technique as instructed, including incorrect use of the inhaler and unsatisfactory medication adherence. This may result in poor control of asthma and increased risk of asthma attacks. In this study, an innova- tive low-cost solution is proposed to monitor the inhalation flow rate. This consists of an acoustic add-on device that generates a sound when using the inhaler. This sound is correlated to the flow rate of the inhalation. The generated sound signal is captured by a smart phone and its features are extracted in order to estimate pa- rameters such as the inspiratory flow rate (IFR), inspiratory volume over the first second (FIV1), total inspiratory lung volume or the inhalation capacity (IC) and other relevant inhalation parameters. Prior to feature extraction, the signal is first passed through a bandpass FIR filter, and then the inhalation segments are detected using a Bayesian sequential detection algorithm. The energy of the pre-processed signal is then used to estimate the inhalation parameters using linear regression. The proposed method is shown to have good performance (e.g. for IFR estimation: (R2 = 99%, p-value< 0.0001).