EUVIMEDEUVIMED entdecken
Crossref

Direct and quantitative lung aeration mapping from raw ultrasound radio-frequency data using Fourier neural operators

Jiayun Wang, Oleksii Ostras, Masashi Sode, Bahareh Tolooshams, Zongyi Li, Kamyar Azizzadenesheli, Gianmarco Pinton, Anima Anandkumar

The Journal of the Acoustical Society of America · 01.10.2025 · 158 (4_Supplement) · A104-A105

DOI: 10.1121/10.0040274 ↗

Kurz gefasst

Fachartikel in der Zeitschrift The Journal of the Acoustical Society of America, erschienen am 01.10.2025. Verfasst von Jiayun Wang und 7 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The physics of lung ultrasound is complex due to reverberation artifacts at the pleural interface, making image interpretation challenging. We hypothesize that first principles of ultrasound wave propagation allow extraction of key diagnostic parameters from wave–tissue interactions using realistic simulations and deep learning. If successful, this would improve diagnostic assessment and enable ultrasound training based on wave physics rather than B-modes. We introduce LUNA, the Lung Ultrasound Neural operator for Aeration, which reconstructs quantitative lung aeration maps directly from raw radio-frequency (RF) ultrasound data. LUNA bypasses conventional beamforming and qualitative B-mode scoring by employing a Fourier neural operator (FNO) that processes the RF signals efficiently in the Fourier domain. The network was trained on 10 150 pairs of simulated RF-aeration pairs, created using a validated finite-difference time-domain model. Afterward, LUNA was fine-tuned with 18 ex-vivo swine lung scans. When evaluated on a set of 103 ex-vivo swine lung samples, LUNA achieved a robust aeration estimation error of 9.4%. For synthetic datasets with 2900 samples, the error was 4.2%. These findings show that lung aeration maps can be reconstructed directly from RF data, laying the groundwork for enhanced interpretability, reproducibility, and diagnostic value in lung ultrasound.

Zitieren

APA

Wang, J., Ostras, O., Sode, M., Tolooshams, B., Li, Z., Azizzadenesheli, K., Pinton, G., & Anandkumar, A. (2025). Direct and quantitative lung aeration mapping from raw ultrasound radio-frequency data using Fourier neural operators. The Journal of the Acoustical Society of America, 158(4_Supplement), A104-A105. https://doi.org/10.1121/10.0040274

Vancouver

Wang J, Ostras O, Sode M, Tolooshams B, Li Z, Azizzadenesheli K, et al. Direct and quantitative lung aeration mapping from raw ultrasound radio-frequency data using Fourier neural operators. The Journal of the Acoustical Society of America. 2025;158(4_Supplement):A104-A105. doi:10.1121/10.0040274

BibTeX
@article{Wang2025,
  author = {Wang, Jiayun and Ostras, Oleksii and Sode, Masashi and Tolooshams, Bahareh and Li, Zongyi and Azizzadenesheli, Kamyar and Pinton, Gianmarco and Anandkumar, Anima},
  title = {Direct and quantitative lung aeration mapping from raw ultrasound radio-frequency data using Fourier neural operators},
  journal = {The Journal of the Acoustical Society of America},
  year = {2025},
  volume = {158},
  number = {4_Supplement},
  pages = {A104--A105},
  doi = {10.1121/10.0040274}
}
Kostenlos starten →