Patient Detection Using Inhalations or Exhalations in a Series of Respirations

Patients with pulmonary diseases exhibit abnormal lung sounds, from which many diseases can be detected through auscultation. If abnormal detection through auscultation can be automated, it will facilitate health management at home for early detection of diseases. We aim to achieve highly accurate automatic detection of pulmonary disease patients using lung auscultation sounds. Lung sounds contain several noises, and for the detection of pulmonary disease patients, it is difficult to rely solely on detecting a single abnormal sound. Instead, a series of respiratory sounds need to be detected. We detected patients using the likelihood of multiple exhalations and inhalations, but the classification rate was low (86.4 %). Abnormal sounds occur only during exhalation or inhalation, and using the likelihood of both in the identification series may contribute to low clas-sification rates. This paper proposes a method for accurately detecting patients by performing patient detection separately for exhalation and inhalation. The experimental results revealed an improvement in the classification rate (89.0 %), indicating the effectiveness of the proposed method.