Full-Waveform Inversion for Breast Ultrasound

U. Taskin

Research output: ThesisDissertation (TU Delft)

51 Downloads (Pure)

Abstract

Breast cancer is the most common type of cancer for women and in developed countries it forms one of their largest threats. Many studies have shown that early detection by screening is important for achieving a successful treatment and reducing the mortality rate. Nowadaysmammography is the gold standard for breast cancer screening. However, mammography has several drawbacks including the use of ionizing radiation, a painful procedure, and poor performance with dense breasts. Magnetic resonance imaging (MRI) could form an alternative as it has some powerful features. However, the high examination and equipment costs as well as the use of contrast agents limits its applicability. Another potential alternative for breast cancer screening is ultrasound. Ultrasound has the advantage over mammography orMRI that it is safe, cheap and patient-friendly. With ultrasound, a tumor can be detected since healthy breast tissues and cancerous tissues have different acoustic properties. All these features make ultrasound a promising candidate as a screening modality for breast cancer. Hand-held ultrasound scans are frequently used for breast imaging in hospitals. With these scanners reflectivity images are generated. These images typically show the boundaries between different tissues. Even when these exams are conducted by trained radiologists operator-dependency occurs. To eliminate this, automated full-breast ultrasound scanners have been developed where the transducer slides over the breast. However, as the imaging principle remains the same, only reflectivity images are generated. To avoid significant breast deformation as well as to scan the breast from as many sides as possible water-bath scanning systems have been developed. These systems have the additional advantage that both reflection and transmission measurements are obtained. This mixture of different measurement types make it feasible to obtain better images by employing advanced processing techniques. One promising imaging method is full-waveform inversion (FWI). FWI aims to match a modeled wavefield to a measured wavefield by adjusting the acoustic medium parameters. A minimization problem is constructed and solved to this aim. As a result, images showing quantitative information about the different tissues are obtained. This quantitative information aids to the characterization and identification of the different tissues. However, there are some challenges when applying FWI. One of the biggest challenges is its computational complexity. By the inclusion of wave phenomena such as diffraction, refraction, scattering and dispersion - needed to explain the measured data in great detail - the computational complexity of FWI has become significantly larger than conventional - mainly ray based - imaging methods. In this work, we investigate the applicability of contrast source inversion (CSI) as an FWI method for breast ultrasound. To this end, we first introduce our full-waveform forward modeling method which is based on solving an integral equation. With a synthetic example,we investigate howeach mediumparameter (compressibility, density, and attenuation) affects the scattered pressure field. The obtained results show that attenuation, in contrast to compressibility and density, has only little effect on the wavefield for frequencies below 1MHz. From that we conclude, that for these frequencies only attenuation can be neglected in our inversion. We also compare the results from our full-waveform modeling method with results obtained after commonly made approximations such as Born, ray-based and paraxial approximations. We observe from the presented numerical results that with each approximation important phenomena normally present in the full-wave data are absent. For this reason, we recommend to use a full-wave modeling method to compute synthetic measurement data.
Original languageEnglish
QualificationDoctor of Philosophy
Awarding Institution
  • Delft University of Technology
Supervisors/Advisors
  • de Jong, N., Supervisor
  • Verschuur, D.J., Supervisor
  • van Dongen, K.W.A., Advisor
Award date26 Apr 2021
Print ISBNs978-94-6384-211-2
DOIs
Publication statusPublished - 2021

Fingerprint

Dive into the research topics of 'Full-Waveform Inversion for Breast Ultrasound'. Together they form a unique fingerprint.

Cite this