Multi-channel lung sound classification with convolutional recurrent neural networks

Messner, Elmar; Fediuk, Melanie; Swatek, Paul; Scheidl, Stefan; Smolle‐Jüttner, Freyja‐Maria; Olschewski, Horst; Pernkopf, Franz

doi:10.1016/j.compbiomed.2020.103831

Cited by 60 publications

(26 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Elmar et al [39] presented an approach for multi-channel lung sound classification using spectral, temporal, and spatial information. They proposed a convolutional recurrent neural network (CRNN) using spectrogram features to classify lung sounds collected from 16 channel recording devices.…”

Section: Lung Disease Classificationmentioning

confidence: 99%

Feature-Based Fusion Using CNN for Lung and Heart Sound Classification

Tariq

Shah

Lee

2022

Sensors

View full text Add to dashboard Cite

Lung or heart sound classification is challenging due to the complex nature of audio data, its dynamic properties of time, and frequency domains. It is also very difficult to detect lung or heart conditions with small amounts of data or unbalanced and high noise in data. Furthermore, the quality of data is a considerable pitfall for improving the performance of deep learning. In this paper, we propose a novel feature-based fusion network called FDC-FS for classifying heart and lung sounds. The FDC-FS framework aims to effectively transfer learning from three different deep neural network models built from audio datasets. The innovation of the proposed transfer learning relies on the transformation from audio data to image vectors and from three specific models to one fused model that would be more suitable for deep learning. We used two publicly available datasets for this study, i.e., lung sound data from ICHBI 2017 challenge and heart challenge data. We applied data augmentation techniques, such as noise distortion, pitch shift, and time stretching, dealing with some data issues in these datasets. Importantly, we extracted three unique features from the audio samples, i.e., Spectrogram, MFCC, and Chromagram. Finally, we built a fusion of three optimal convolutional neural network models by feeding the image feature vectors transformed from audio features. We confirmed the superiority of the proposed fusion model compared to the state-of-the-art works. The highest accuracy we achieved with FDC-FS is 99.1% with Spectrogram-based lung sound classification while 97% for Spectrogram and Chromagram based heart sound classification.

show abstract

Section: Lung Disease Classificationmentioning

confidence: 99%

Feature-Based Fusion Using CNN for Lung and Heart Sound Classification

Tariq

Shah

Lee

2022

Sensors

View full text Add to dashboard Cite

show abstract

“…The multi-channel lung sound database [9], [19], [28] has been recorded in a clinical trial. It contains lung sounds of 16 healthy subjects and 7 patients diagnosed with idiopathic pulmonary fibrosis (IPF).…”

Section: B Multi-channel Lung Sound Databasementioning

confidence: 99%

“…In [28], a classification framework using lung sound signals of all recording channels was introduced to identify healthy and pathological breathing cycles. Lung sounds of one breath cycle of all recording channels were first transformed into STFT spectrograms.…”

Section: B Lung Sound Classification On Our Multi-channel Datasetmentioning

confidence: 99%

Lung Sound Classification Using Co-tuning and Stochastic Normalization

Nguyen¹,

Pernkopf²

2021

Preprint

Self Cite

View full text Add to dashboard Cite

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 pretrained model is transferred by using vanilla fine-tuning, cotuning, stochastic normalization and the combination of the cotuning 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.

show abstract

“…The training target of SVM is shown as (20) A Lagrangian is selected for optimization: (21) By setting the derivatives of L to zero with respect to ω and b, ω is obtained as follows: (22) The training target is reformulated as follows: (23) To solve the non-separable case, the regularization factors C are introduced and reformulated Eq. ( 23): (24) To reduce the operational complexity of the inner products, the kernel functions are used to replace the inner product: (25) The regular method used to obtain the coefficients is the sequential minimal optimization (SMO) algorithm [39]. For the SVM parameters, a context-aware support vector machine (C-SVM) is selected as the SVM type and a radial basis function (RBF) as the kernel function.…”

Section: A Support Vector Machinementioning

confidence: 99%

“…Jaber M. M et al [23] proposed a telemedicine framework for lung sound based on the telemedicine framework. Messner E et al [24] proposed a multi-channel lung sound classification method. They selected the convolutional recurrent neural network and obtained the F1 score approximates to 92%.…”

Section: Introductionmentioning

confidence: 99%

Detection of Respiratory Sounds Based on Wavelet Coefficients and Machine Learning

et al. 2020

View full text Add to dashboard Cite

Respiratory sounds reveal important information of the lungs of patients. However, the analysis of lung sounds depends significantly on the medical skills and diagnostic experience of the physicians and is a time-consuming process. The development of an automatic respiratory sound classification system based on machine learning would, therefore, be beneficial. In this study, 705 respiratory sound signals (240 crackles, 260 rhonchi, and 205 normal respiratory sounds) were acquired from 130 patients. We found that similarities between the original and wavelet decomposed signals reflected the frequency of the signals. The Gaussian kernel function was used to evaluate the wavelet signal similarity. We combined the wavelet signal similarity with the relative wavelet energy and wavelet entropy as the feature vector. A 5-fold cross-validation was applied to assess the performance of the system. The artificial neural network model, which was applied, achieved the classification accuracy and classified the respiratory sound signals with an accuracy of 85.43%.

show abstract

Multi-channel lung sound classification with convolutional recurrent neural networks

Cited by 60 publications

References 25 publications

Feature-Based Fusion Using CNN for Lung and Heart Sound Classification

Feature-Based Fusion Using CNN for Lung and Heart Sound Classification

Lung Sound Classification Using Co-tuning and Stochastic Normalization

Detection of Respiratory Sounds Based on Wavelet Coefficients and Machine Learning

Contact Info

Product

Resources

About