Deep Lung Auscultation Using Acoustic Biomarkers for Abnormal Respiratory Sound Event Detection

Abstract

Lung Auscultation is a non-invasive process of distinguishing normal respiratory sounds from abnormal ones by analyzing the airflow along the respiratory tract. With developments in the Deep Learning (DL) techniques and wider access to anonymized medical data, automatic detection of specific sounds such as crackles and wheezes have been gaining popularity. In this paper, we propose to use two sets of diversified acoustic biomarkers extracted using Discrete Wavelet Transform (DWT) and deep encoded features from the intermediate layer of a pre-trained Audio Event Detection (AED) model trained using sounds from daily activities. First set of biomarkers highlight the time frequency localization characteristics obtained from DWT coefficients. However, the second set of deep encoded biomarkers captures a generalized reliable representation, and thus indemnifies the scarcity of training samples and the class imbalance in dataset. The model trained using these features achieves a 15.05% increase in terms of the specificity over the baseline model that uses spectrogram features. Moreover, ensemble of DWT features and deep encoded feature based models show absolute improvements of 8.32%, 6.66% and 7.40% in terms of sensitivity, specificity and ICBHIscore, respectively, and clearly outperforms the state-of-the-art with a significant margin.

Compressed Sensing Data with Performing Audio Signal Reconstruction for the Intelligent Classification of Chronic Respiratory Diseases

“…[ 39 ] utilized discrete wavelet transforms and deep learning for classifying the ICBHI 2017 challenge database into healthy and unhealthy, which achieved an F1 score of 81.64%, similar to the F1 scores of the best models of Healthy versus COPD of 83%. However, this study’s approach was more focused on COPD, whereas [ 39 ] unhealthy had a broader range of diseases. Ref.…”

Exploring machine learning for audio-based respiratory condition screening: A concise review of databases, methods, and open issues

“…RNN can also overcome the restricted visual field of CNNs, leading to better cross-time and long-distance correlation modeling. Tiwari et al 59 developed a bi-directional RNN model via ICBHI 2017 database, yielding an accuracy of over 80% in detecting abnormal respiratory cycles. RNN can also be jointly applied with CNN model to better capture spatial–temporal features for respiratory sound classification.…”

“…Compared to feature-based ML models, deep learning models do not depend on explicit feature engineering, so they usually present more powerful capability of modeling audio-disease relations with the premise of massive training data. The latest state-of-the-art audio-based respiratory condition screening methods are mainly deep learning based, covering convolutional neural networks (CNNs), 32 , 61 recurrent neural networks (RNNs), 59 , 60 and Transformer neural networks. 41 , 79 Those models have demonstrated favorable performance in detecting COPD, asthma, and other respiratory conditions.…”