Lung Sound Classification Using Co-tuning and Stochastic Normalization

TL;DR

This study enhances lung sound classification by leveraging pre-trained ResNet models with co-tuning and stochastic normalization, combined with data augmentation and spectrum correction, achieving superior performance over existing methods.

Contribution

Introduces a novel combination of co-tuning and stochastic normalization techniques with data augmentation for improved lung sound classification.

Findings

Outperforms state-of-the-art systems on lung sound datasets.

Effective handling of class imbalance through data augmentation.

Improved robustness with spectrum correction.

Abstract

In this paper, we use pre-trained ResNet models as backbone architectures for classification of adventitious lung sounds and respiratory diseases. The knowledge of the pre-trained model is transferred by using vanilla fine-tuning, co-tuning, stochastic normalization and the combination of the co-tuning and stochastic normalization techniques. Furthermore, data augmentation in both time domain and time-frequency domain is used to account for the class imbalance of the ICBHI and our multi-channel lung sound dataset. Additionally, we apply spectrum correction to consider the variations of the recording device properties on the ICBHI dataset. Empirically, our proposed systems mostly outperform all state-of-the-art lung sound classification systems for the adventitious lung sounds and respiratory diseases of both datasets.