Reconstruction of Transmission Images by Full Waveform Inversion (FWI) for 3D Ultrasound Computer Tomography
by
IPE Bldg. 242, Seminarroom EG + Zoom
Final presentation of Nam's master thesis
Abstract:
This thesis explores the use of frequency-domain full waveform inversion (FWI) for quantitative ultrasound computer tomography (USCT) imaging, with a focus on reconstructing speed-of-sound and attenuation maps. Starting from a 2D ring-array approximation, the method is adapted to process experimental and simulated data from the semi-elliptical 3D USCT III system. A complete FWI pipeline is developed, including preprocessing, transmission signal extraction, frequency-domain conversion, and model initialization using straight-ray tomography. The study investigates how well 2D reconstructions can represent 3D measurements, and evaluates the influence of transducer arrangement, noise levels, and initial model choice. The approach is then extended to 3D reconstruction using a frequency-domain Helmholtz solver. Results on simulated, phantom, and patient data show that the adapted 2D method can produce meaningful reconstructions from 3D acquisitions, while maintaining a relative low computational effort compared to 3D reconstructions. This work provides a practical bridge between 2D and 3D FWI in USCT, highlighting both the opportunities and limitations for clinical breast imaging.
Timo Muscheid