Home | english  | Fakultät | Impressum | Datenschutz | KIT

Presentation of the Master's Thesis: Ultrasound medical imaging using 2D viscoacoustic full-waveform inversion

Presentation of the Master's Thesis: Ultrasound medical imaging using 2D viscoacoustic full-waveform inversion



Fabian Kühn


In the last 20 years full-waveform inversion (FWI) has has been developed in the geophysical community with great success, as it is able to provide multi-parameter models in high resolution. In FWI, an initial model is iteratively adjusted by minimizing the misfit between the synthetic and the measured data using e.g. adjoint techniques. By exploiting more information of the measured waveform compared to traveltime tomography, FWI is able to exceed the resolution limit of ray-based methods at the expense of much higher computational costs.

In this study we want to investigate the benefits of transferring the well established seismic FWI technology to ultrasound medical imaging, especially breast cancer screening. Nowadays, mammography using X-rays, is the common approach for breast cancer screening. However, it is more and more often replaced by ultrasound tomography, which is non-invasive and provides more reproducible results due to the fact that the breast is not deformed during data acquisition.

In contrast to the seismic application of FWI, elastic parameters have a negligible effect on wave propagation in breast tissue, which allows us to use the acoustic approximation of the wave equation for wavefield simulations.

In this work, we perform two-dimensional synthetic reconstruction tests using a realistic numerical breast model. The sound of speed within different breast tissues varies between approximately 1400 and 1700 meters per second. We use source signals with frequencies up to a few megahertz, which lead to wavelengths in the order of millimeters. Considering a transmission geometry, where sources and receivers are located all around the breast in a ring array having a diameter of 20 cm, we have to propagate about 100 to 150 wavelength in the target.

Furthermore, we analyse a clinical ultrasound dataset from the KIT-IPE 3D USCT project to assess the applicability of FWI on datasets acquired with their ultrasound device and to investigate the related challenges.

We study the potential of FWI to help improving the quality of speed of sound reconstruction of the breast. Increased breast density is a well-known breast cancer risk factor and complicates imaging of tumors. Improvements in ultrasound tomography based on FWI could help to detect small cancers earlier, in particular among the aforementioned risk group, and thus improve the survival probability of patients.