Hybrid deep learning and YAMNet features for asthma diagnosis from respiratory sounds
Clinical Snapshot
PICO Framework
| P — Population | Patients with respiratory sounds from the Asthma Detection Dataset v2, comprising actual audio recordings of respiratory noises |
| I — Intervention | Hybrid deep learning architecture combining YAMNet embeddings with handcrafted acoustic descriptors (MFCCs, chromagram, ZCR, spectral features) |
| C — Comparator | Not explicitly stated - appears to be compared against baseline performance or other machine learning approaches |
| O — Outcomes | Diagnostic accuracy for asthma detection (accuracy, F1-score, macro-AUC, sensitivity, specificity) |
Bottom Line
This study presents a sophisticated machine learning approach for asthma diagnosis using respiratory sounds, achieving impressive accuracy metrics (98.6%) on a public dataset. However, significant limitations include lack of external validation, unclear reference standards, and no comparison with current clinical diagnostic methods. While the technical approach is sound and includes interpretability features, the clinical utility remains unproven. The extremely high performance metrics raise concerns about overfitting and generalizability to real-world clinical settings. Before clinical implementation, this technology requires validation against gold-standard spirometry and testing in diverse patient populations with varying disease severity and comorbidities. The approach shows promise as a screening tool, particularly in resource-limited settings, but cannot replace comprehensive clinical assessment and established diagnostic methods.
Key Findings
P Value: Not reported
Effect Size: Accuracy: 98.6% ± 0.14%
Primary Outcome: Asthma diagnostic accuracy using respiratory sound analysis
Nnt Or Sensitivity: Sensitivity and specificity not explicitly reported, though high AUC suggests excellent discrimination
Confidence Interval: F1-score: 97.5% ± 0.14%, macro-AUC: 99.1% ± 0.14%
Clinical Application
Requires smartphone or digital recording device and computational resources for model deployment Could complement PBS-subsidised spirometry and support RACGP guidelines for asthma diagnosis, particularly in remote areas with limited specialist access Patients with suspected asthma requiring diagnostic evaluation, particularly in resource-limited settings
Abstract
A prompt and precise diagnosis is necessary to lessen the worldwide burden of asthma, a common chronic respiratory condition. Spirometry is one of the resource-intensive conventional methods that frequently misclassifies cases because of overlapping symptoms with other respiratory illnesses. The Asthma Detection Dataset v2, which is openly accessible and comprises actual audio recordings of respiratory noises, is used in this work to demonstrate a hybrid deep learning architecture for automated asthma detection. After applying thorough preprocessing, which includes denoising, segmentation, and augmentation, the suggested pipeline employs a dual-path feature extraction technique that combines pretrained YAMNet embeddings with handmade acoustic descriptors (MFCCs, chromagram, ZCR, and spectral features). Atrous Spatial Pyramid Pooling (ASPP) and Squeeze-and-Excitation (SE) modules are used to further improve these embeddings, and a Multi-Layer Perceptron is used for classification. The proposed model achieved high performance, with an accuracy of 98.6% ± 0.14, an F1-score of 97.5% ± 0.14, and a macro-AUC of 99.1% ± 0.14 using stratified 5-fold cross-validation. SHAP-based interpretability and visualization (waveforms, spectrograms, t-SNE) verified that the model records clinically significant auditory biomarkers such as crackles and wheezes. These results indicate the potential of the proposed model as an explainable and scalable computer-aided asthma diagnostic tool.
References
- 1.Shatat, G., Moustafa, H., Saraya, M. S., Marzouk, A. A., & Hekal, A. A. (2026). Hybrid deep learning and YAMNet features for asthma diagnosis from respiratory sounds. Scientific Reports. https://doi.org/10.3390/electronics14142794
Related Research
Sleep medicine
Concordance of wearable device sleep metrics with patient-reported sleep quality: A systematic review
3 Aug 2026
Bioelectrochemistry (Amsterdam, Netherlands)
From data to diagnosis: A comprehensive review of machine learning-driven wearable sensors in healthcare.
2 Aug 2026
JMIR mHealth and uHealth
Wearable Devices for Monitoring and Management of Comorbid Obstructive Sleep Apnea and Hypertension: Scoping Review
1 Aug 2026
This content is for educational purposes for healthcare professionals only and does not constitute clinical advice. Clinical decisions should be based on individual patient assessment, current guidelines, and appropriate specialist consultation. Editorial Standards · Privacy Policy · Terms of Service