Lung disease classification detection method based on collaborative deep learning and lung auscultation sound

Abstract

The invention provides a lung disease classification detection method based on collaborative deep learning and lung auscultation sound, and the method employs a method of collaborative learning between two ResNet-50, carries out training through a pairwise learning mode, each time a model receives a data pair as input, and a pair of audio feature values are respectively transmitted to the corresponding ResNet-50; the ResNet-50 is initialized and trained by adopting a method of fine tuning of a pre-training model, a collaborative learning system is designed, the two ResNet-50 are enabled to perform mutual-aid learning, and the collaborative system is used for supervising similarities and differences attributes of data pairs, namely whether the data pairs belong to one category or not. According to the method, two ResNet-50 collaborative errors are acquired, the collaborative errors generated by the two ResNet-50 are subjected to back propagation in real time, and the weight of the network is corrected, so that the feature representation learning capability of the network is further enhanced, and easy-to-confuse samples can be more effectively and accurately distinguished.

Description

technical field [0001] The invention belongs to the technical field of classification and detection of medical data, and in particular relates to a method for classification and detection of lung diseases based on collaborative deep learning and lung auscultation sounds. Background technique [0002] The lung is a vital organ to maintain the normal operation of the body. It communicates with the outside world. There are 10,000 liters of gas in and out every day. Activity also plays a big role in the body's metabolism and immune function. However, since entering the industrial society, with the continuous development of industrialization and economy, its by-products have also caused certain damage to the environment, and the number of patients suffering from respiratory diseases is increasing. Factors such as the ubiquitous population aging are closely related. Nowadays, respiratory system diseases have become a common and frequently-occurring disease. Most of the lesions a...

AI Technical Summary

Problems solved by technology

Doctors can judge the type of disease through different characteristics of breath sounds such as alveolar breath sounds, bronchial sounds, and chest friction sounds. It has a good effect on early detection of lung lesions. It is not completely within the sensitive frequency range of the human ear, so it is easy to lose important low-frequency information during auscultation of the human ear, which affects the accuracy and reliability of the diagnostic results

Moreover, the effect of auscultation is closely related to the doctor's level and experience, and is also affected by the external environment and hearing conditions. For various reasons, there may be a risk of missed diagnosis and misdiagnosis of lung diseases relying on lung auscultation sounds.

Images