Lung sound diagnosis device based on deep learning

Abstract

The invention discloses a lung sound diagnosis device based on deep learning. The lung sound diagnosis device comprises acquisition equipment, an image labeling tool, a data preprocessing module, a secondary data enhancement module and a deep learning classification model which are connected in sequence; the acquisition equipment is used for respectively acquiring lung auscultation data of normal and abnormal lung health conditions; the image annotation tool is used for dividing and annotating each section of collected audio; the data preprocessing module and the secondary data enhancement module are used for preprocessing the collected audio signal samples, performing data enhancement to obtain audio primary features of different lung auscultation signal samples, classifying and marking the audio primary features as normal samples and abnormal samples for subsequent further deep learning feature extraction; and the deep learning classification model is used for performing deep learning on a classification model, performing training according to the primary feature vector to obtain a series of high-level features, performing classification by adopting an RF classifier, and obtaining a multi-classification RF model through the input high-level features.

Description

technical field [0001] The invention relates to the field of smart medical technology, in particular to an audio signal classification device based on a deep learning model. Background technique [0002] The lung is the respiratory organ of the human body and the main place for gas exchange between the human body and the outside world. In the process of gas exchange, the lungs produce different sounds, such as low-pitched dry rales, large blisters, medium blisters, and small blisters. These sounds indirectly reflect the health problems of the lungs and airways: low-pitched dry rales correspond to abnormal trachea or main trachea; Pneumonia, small blisters sound corresponding to bronchiolitis, early pulmonary congestion. In the traditional medical field, doctors need to use a stethoscope to listen to the audio and rely on their own experience to judge the patient's condition. This artificial method is inefficient and relies too much on the doctor's personal experience, and ...

AI Technical Summary

Problems solved by technology

In the traditional medical field, doctors need to use a stethoscope to listen to the audio and rely on their own experience to judge the patient's condition. This manual method is inefficient and relies too much on the doctor's personal experience, and the accuracy rate is not stable; Some methods based on deep learning only use some relatively basic network models, which cannot extract features very well, and the effect is not good

Images