EUVIMEDEUVIMED entdecken
Crossref

Multitask learning-based phonocardiogram denoising model for preserving valvular heart disease characteristics

Geon Lee, Hangsik Shin iD

Physiological Measurement · 31.07.2026 · 47 (7) · 075017

DOI: 10.1088/1361-6579/ae8caa ↗

Kurz gefasst

Fachartikel in der Zeitschrift Physiological Measurement, online erschienen am 31.07.2026. Verfasst von Geon Lee und 1 weiteren Autor:innen. Beteiligte Einrichtung: Department of Medical Informatics and Statistics, Brain Korea 21 Project, Asan Medical Center, University of Ulsan College of Medicine, Seoul 05505, …. Hauptthemen laut MEDLINE: Signal-To-Noise Ratio, Signal Processing, Computer-Assisted, Heart Valve Diseases, Machine Learning. Die Arbeit nennt 38 Literaturangaben. Gefördert von: the Ministry of Science and ICT, the Ministry of Trade, Industry and Energy, the Ministry of Health & Welfare, the Ministry of Food and Drug Safety, Ministry of Trade Industry & Energy.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Objective. This study developed a multitask learning (MTL)-based denoising model to reconstruct phonocardiogram (PCG) signals while preserving heart murmur characteristics under both synthetic and diverse real-world clinical noise. Approach. The model was trained to jointly denoise PCG data and classify valvular heart disease (VHD). The original PCG dataset comprised recordings lasting up to 3 s, including normal heart sounds and four VHD classes, with 200 samples per class. To construct inputs, we synthesized contaminated PCGs by mixing synthetic and real-world clinical noise sources and converted them into spectrogram images. The model was evaluated at signal-to-noise ratios (SNRs) of −5, 0, 5, and 10 dB using 5 × 5 nested cross-validation. Denoising performance was measured by the scale-invariant signal-to-distortion ratio (SI-SDR), and VHD classification by accuracy. Main results. Our model outperformed a single-task U-Net denoising baseline; even at a highly contaminated SNR of −5 dB, it achieved an absolute mean SI-SDR of 14.05 dB. For real-world noise, including hospital ambient noise and lung sounds, our model improved SI-SDR by up to 9.69 dB over previous work. For VHD classification, the model exceeded 98.5% accuracy across most conditions at SNRs of 5 and 10 dB, suggesting that heart murmur-specific features were preserved during denoising. Significance. Overall, this MTL-based model demonstrated consistent and robust PCG reconstruction across diverse noise conditions, improving VHD classification on contaminated PCG signals.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Signal-To-Noise RatioSignal Processing, Computer-AssistedHeart Valve DiseasesMachine LearningPhonocardiographyHumans

Stichwörter der Autor:innen

deep learningmultitask learningnoise reductionphonocardiogramvalvular heart disease classification

Autor:innen und Institutionen

  • Department of Medical Informatics and Statistics, Brain Korea 21 Project, Asan Medical Center, University of Ulsan College of Medicine, Seoul 05505, Republic of Korea.
  • Institute of Digital Healthcare, University of Ulsan College of Medicine, Seoul 05505, Republic of Korea.

ORCID-Profile: Hangsik Shin

Förderung

  • the Ministry of Science and ICT, the Ministry of Trade, Industry and Energy, the Ministry of Health & Welfare, the Ministry of Food and Drug Safety
  • Ministry of Trade Industry & Energy

Zitieren

APA

Lee, G., & Shin, H. (2026). Multitask learning-based phonocardiogram denoising model for preserving valvular heart disease characteristics. Physiological Measurement, 47(7), 075017. https://doi.org/10.1088/1361-6579/ae8caa

Vancouver

Lee G, Shin H. Multitask learning-based phonocardiogram denoising model for preserving valvular heart disease characteristics. Physiological Measurement. 2026;47(7):075017. doi:10.1088/1361-6579/ae8caa

BibTeX
@article{Lee2026,
  author = {Lee, Geon and Shin, Hangsik},
  title = {Multitask learning-based phonocardiogram denoising model for preserving valvular heart disease characteristics},
  journal = {Physiological Measurement},
  year = {2026},
  volume = {47},
  number = {7},
  pages = {075017},
  doi = {10.1088/1361-6579/ae8caa}
}
Kostenlos starten →