Background
This project is a collaboration between CWI (Dr. Felix Lucka, Prof. Tristan van Leeuwen), the University of Technology Eindhoven (Dr. Hans-Martin Schwab), and Tomosono (Dr.ir. Gert Schoonenberg).
Motivation
At Tomosono and TU/e, a new prototype scanner for 3D Ultrasound Tomography (UST) is being build and tested. It consists of multiple panels with dedicated capacitive micromachined ultrasonic transducer (CMUT) arrays arranged in a ring to obtain a cylindrical field of view. In this project, we take the first steps to develop a model-based UST image reconstruction technique called Full-Waveform Inversion (FWI) for this scanner. We will work with simulated, idealized data, only.
Outline
FWI relies on a numerical wave propagation model for which we will use the open source MATLAB toolbox k-Wave, which implements a time-domain (TD) k-space pseudospectral method. Many properties of the scanner will be idealized in this project, for example, we will assume to know the exact position of the transducers, and that they have ideal spatial and temporal responses. We will design a simple numerical phantom to simulate data and evaluate the quality of the reconstructed images. We will set up and evaluate the key components of TD-FWI schemes, including the stochastic optimization scheme, the stochastic gradient estimator, the multigrid strategy, step sizes and iteration numbers. At the end of the project, we will use the developed methods to examine how changes in the data acquisition design impact the image quality.
