Classification of lung sounds in patients with asthma, emphysema, fibrosing alveolitis and healthy lungs by using self-organizing maps

Deep learning diagnostic and severity-stratification for interstitial lung diseases and chronic obstructive pulmonary disease in digital lung auscultations and ultrasonography: clinical protocol for an observational case-control study

BACKGROUND

Interstitial lung diseases (ILD), such as idiopathic pulmonary fibrosis (IPF) and non-specific interstitial pneumonia (NSIP), and chronic obstructive pulmonary disease (COPD) are severe, progressive pulmonary disorders with a poor prognosis. Prompt and accurate diagnosis is important to enable patients to receive appropriate care at the earliest possible stage to delay disease progression and prolong survival. Artificial intelligence-assisted lung auscultation and ultrasound (LUS) could constitute an alternative to conventional, subjective, operator-related methods for the accurate and earlier diagnosis of these diseases. This protocol describes the standardised collection of digitally-acquired lung sounds and LUS images of adult outpatients with IPF, NSIP or COPD and a deep learning diagnostic and severity-stratification approach.

METHODS

A total of 120 consecutive patients (≥ 18 years) meeting international criteria for IPF, NSIP or COPD and 40 age-matched controls will be recruited in a Swiss pulmonology outpatient clinic, starting from August 2022. At inclusion, demographic and clinical data will be collected. Lung auscultation will be recorded with a digital stethoscope at 10 thoracic sites in each patient and LUS images using a standard point-of-care device will be acquired at the same sites. A deep learning algorithm (DeepBreath) using convolutional neural networks, long short-term memory models, and transformer architectures will be trained on these audio recordings and LUS images to derive an automated diagnostic tool. The primary outcome is the diagnosis of ILD versus control subjects or COPD. Secondary outcomes are the clinical, functional and radiological characteristics of IPF, NSIP and COPD diagnosis. Quality of life will be measured with dedicated questionnaires. Based on previous work to distinguish normal and pathological lung sounds, we estimate to achieve convergence with an area under the receiver operating characteristic curve of > 80% using 40 patients in each category, yielding a sample size calculation of 80 ILD (40 IPF, 40 NSIP), 40 COPD, and 40 controls.

DISCUSSION

This approach has a broad potential to better guide care management by exploring the synergistic value of several point-of-care-tests for the automated detection and differential diagnosis of ILD and COPD and to estimate severity. Trial registration Registration: August 8, 2022.

CLINICALTRIALS

gov Identifier: NCT05318599.

Lung Sound Classification Using Co-tuning and Stochastic Normalization

Computational methods for lung sound analysis are beneficial for computer-aided diagnosis support, storage and monitoring in critical care. In this paper, we use pre-trained ResNet models as backbone architectures for classification of adventitious lung sounds and respiratory diseases. The learned representation 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 introduce spectrum correction to account for 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.

Lung Sound Classification Using Snapshot Ensemble of Convolutional Neural Networks

We propose a robust and efficient lung sound classification system using a snapshot ensemble of convolutional neural networks (CNNs). A robust CNN architecture is used to extract high-level features from log mel spectrograms. The CNN architecture is trained on a cosine cycle learning rate schedule. Capturing the best model of each training cycle allows to obtain multiple models settled on various local optima from cycle to cycle at the cost of training a single mode. Therefore, the snapshot ensemble boosts performance of the proposed system while keeping the drawback of expensive training of ensembles moderate. To deal with the class-imbalance of the dataset, temporal stretching and vocal tract length perturbation (VTLP) for data augmentation and the focal loss objective are used. Empirically, our system outperforms state-of-the-art systems for the prediction task of four classes (normal, crackles, wheezes, and both crackles and wheezes) and two classes (normal and abnormal (i.e. crackles, wheezes, and both crackles and wheezes)) and achieves 78.4% and 83.7% ICBHI specific micro-averaged accuracy, respectively. The average accuracy is repeated on ten random splittings of 80% training and 20% testing data using the ICBHI 2017 dataset of respiratory cycles.

Lung sound analysis helps localize airway inflammation in patients with bronchial asthma

Purpose

Airway inflammation can be detected by lung sound analysis (LSA) at a single point in the posterior lower lung field. We performed LSA at 7 points to examine whether the technique could identify the location of airway inflammation in patients with asthma.

Patients and methods

Breath sounds were recorded at 7 points on the body surface of 22 asthmatic subjects. Inspiration sound pressure level (I_SPL), expiration sound pressure level (E_SPL), and the expiration-to-inspiration sound pressure ratio (E/I) were calculated in 6 frequency bands. The data were analyzed for potential correlation with spirometry, airway hyperresponsiveness (PC₂₀), and fractional exhaled nitric oxide (FeNO).

Results

The E/I data in the frequency range of 100–400 Hz (E/I low frequency [LF], E/I mid frequency [MF]) were better correlated with the spirometry, PC₂₀, and FeNO values than were the I_SPL or E_SPL data. The left anterior chest and left posterior lower recording positions were associated with the best correlations (forced expiratory volume in 1 second/forced vital capacity: r=−0.55 and r=−0.58; logPC₂₀: r=−0.46 and r=−0.45; and FeNO: r=0.42 and r=0.46, respectively). The majority of asthmatic subjects with FeNO ≥70 ppb exhibited high E/I MF levels in all lung fields (excluding the trachea) and V₅₀%pred <80%, suggesting inflammation throughout the airway. Asthmatic subjects with FeNO <70 ppb showed high or low E/I MF levels depending on the recording position, indicating uneven airway inflammation.

Conclusion

E/I LF and E/I MF are more useful LSA parameters for evaluating airway inflammation in bronchial asthma; 7-point lung sound recordings could be used to identify sites of local airway inflammation.

Artificial intelligence techniques used in respiratory sound analysis--a systematic review

Artificial intelligence (AI) has recently been established as an alternative method to many conventional methods. The implementation of AI techniques for respiratory sound analysis can assist medical professionals in the diagnosis of lung pathologies. This article highlights the importance of AI techniques in the implementation of computer-based respiratory sound analysis. Articles on computer-based respiratory sound analysis using AI techniques were identified by searches conducted on various electronic resources, such as the IEEE, Springer, Elsevier, PubMed, and ACM digital library databases. Brief descriptions of the types of respiratory sounds and their respective characteristics are provided. We then analyzed each of the previous studies to determine the specific respiratory sounds/pathology analyzed, the number of subjects, the signal processing method used, the AI techniques used, and the performance of the AI technique used in the analysis of respiratory sounds. A detailed description of each of these studies is provided. In conclusion, this article provides recommendations for further advancements in respiratory sound analysis.

Lung sounds in bronchial asthma

Modern understanding of lung sounds started with a historical article by Forgacs. Since then, many studies have clarified the changes of lung sounds due to airway narrowing as well as the mechanism of genesis for these sounds. Studies using bronchoprovocation have shown that an increase of the frequency and/or intensity of lung sounds was a common finding of airway narrowing and correlated well with lung function. Bronchoprovocation studies have also disclosed that wheezing may not be as sensitive as changes in basic lung sounds in acute airway narrowing. A forced expiratory wheeze (FEW) may be an early sign of airway obstruction in patients with bronchial asthma. Studies of FEW showed that airway wall oscillation and vortex shedding in central airways are the most likely mechanisms of the generation of expiratory wheezes. Studies on the genesis of wheezes have disclosed that inspiratory and expiratory wheezes may have the same mechanism of generation as a flutter/flow limitation mechanism, either localized or generalized. In lung sound analysis, the narrower the airways are, the higher the frequency of breathing sounds is, and, if a patient has higher than normal breathing sounds, i.e., bronchial sounds, he or she may have airway narrowing or airway inflammation. It is sometimes difficult to detect subtle changes in lung sounds; therefore, we anticipate that automated analysis of lung sounds will be used to overcome these difficulties in the near future.

Combining neural network and genetic algorithm for prediction of lung sounds

Recognition of lung sounds is an important goal in pulmonary medicine. In this work, we present a study for neural networks-genetic algorithm approach intended to aid in lung sound classification. Lung sound was captured from the chest wall of The subjects with different pulmonary diseases and also from the healthy subjects. Sound intervals with duration of 15-20 s were sampled from subjects. From each interval, full breath cycles were selected. Of each selected breath cycle, a 256-point Fourier Power Spectrum Density (PSD) was calculated. Total of 129 data values calculated by the spectral analysis are selected by genetic algorithm and applied to neural network. Multilayer perceptron (MLP) neural network employing backpropagation training algorithm was used to predict the presence or absence of adventitious sounds (wheeze and crackle). We used genetic algorithms to search for optimal structure and training parameters of neural network for a better predicting of lung sounds. This application resulted in designing of optimum network structure and, hence reducing the processing load and time.

Two-stage classification of respiratory sound patterns

The classification problem of respiratory sound signals has been addressed by taking into account their cyclic nature, and a novel hierarchical decision fusion scheme based on the cooperation of classifiers has been developed. Respiratory signals from three different classes are partitioned into segments, which are later joined to form six different phases of the respiration cycle. Multilayer perceptron classifiers classify the parameterized segments from each phase and decision vectors obtained from different phases are combined using a nonlinear decision combination function to form a final decision on each subject. Furthermore a new regularization scheme is applied to the data to stabilize training and consultation.

Spirometry and forced oscillometry assisted optimal frequency band determination for the computerized analysis of tracheal lung sounds in asthma

We analysed respiration sounds of individual asthmatic patients, in the scope of the development of a method for computerized recognition of the degree of airway obstruction. Respiration sounds were recorded during laboratory sessions of histamine-provoked airway obstruction. We applied an interpolation technique using supervised artificial neural networks to investigate the optimal frequency band required for studying tracheal asthmatic lung sounds. The optimal band was found to be 100-2300 Hz. The forced expiratory volume in 1 s (FEV1) and the respiratory resistance parameter Rrs(4) were used to describe the degree of airway obstruction that is associated with the lung sounds. By comparing the results obtained with the two parameters, we found that for parametrization of the associated degree of airway obstruction respiratory resistance measurements are preferable over forced expiratory volume measurements.

Lung function interpolation by means of neural-network-supported analysis of respiration sounds

Respiration sounds of individual asthmatic patients were analysed in the scope of the development of a method for computerised recognition of the degree of airways obstruction. Respiration sounds were recorded during laboratory sessions of allergen provoked airways obstruction, during several stages of advancing obstruction. The technique of artificial neural networks was applied for relating sound spectra and simultaneously measured lung function values (spirometry parameter FEV(1)). The ability of feedforward neural networks was tested to interpolate obstruction levels of FEV(1)-classes of which no members were included in the set used for training a network. In this way, a situation was simulated of an existing network recognising a new asthmatic attack under the same physiological conditions. It appeared to be possible to interpolate FEV(1) values, and it is concluded that a deterministic relationship exists between sound spectra and lung function parameter FEV(1). Variance optimisation appeared to be important in optimising the neural network configuration.