Browse Teach Curate

Direct Reconstruction of Lung Aeration Maps from Ultrasound RF Data with Fourier Neural Operators

Students learn how a Fourier neural operator can map raw ultrasound radio-frequency data directly to a spatial lung aeration map, skipping traditional beamforming and B-mode interpretation. The approach frames aeration recovery as an ill-posed inverse problem solved in Fourier space, trained on full-wave-simulated data and fine-tuned on real tissue, and shows how a physical quantity (percent aeration) can be estimated reproducibly rather than through operator-dependent artifact reading.

KITT Text

Ultrasound Lung Aeration Map via Physics-Aware Neural Operators Jiayun Wang1∗, Oleksii Ostras2,3∗

This paper introduces Luna, a physics-aware deep-learning model that reconstructs quantitative lung aeration maps directly from raw ultrasound radio-frequency (RF) data, bypassing conventional delay-…