Bachelor and Master Theses

Background

This thesis is set within the research field of Microwave Imaging (MWI) for breast cancer detection, that has been subject to research for more than two decades as a complementary method to already established imaging modalities like X-ray mammography or Ultrasound to improve the reliability of a diagnose.

In MWI, an electromagnetic wave (EMW) is sent out that propagates towards an object-under-study (OUS), here the female breast, where it gets reflected, transmitted, and diffracted. Based on the deflected signal that is detected by one or multiple receiving antennas, information can be retrieved about the nature of the OUS. Therefore, MWI is an inherently inverse problem, i.e., the tissue distribution inside the OUS must be reconstructed from a finite number of measurements taken outside of the OUS. A multitude of image reconstruction algorithms has been presented for that purpose, differing in, e.g., complexity, information content, or signal configuration.

Motivation

One of the most common imaging algorithms applie a model-based approach. For that purpose, an accurate numerical model of the experimental imaging setup is required that incorporates all components of the measurement setup except for the unknown tissue distribution inside the OUS. The simulated tissue distribution is then iteratively updated until a good agreement between simulated and measured signals has been found. The simulated tissue distribution is then assumed to resemble the actual distribution inside the breast.

At the MWI group at MDU, a new transmitting applicator has recently been developed that induces a propagating EMW directly inside a dielectric object by means of a magnetic near-field. A numerical model has been created and validated against experimental measurements with different dielectric loads, but no image reconstruction has been performed so far.

Tasks

In this thesis, the student will:

-       Generate numerical data using a commercially available electromagnetic forward solver with a fully defined simulation setup to create a ground truth.

-       Implement their own electromagnetic forward solver in which the unknown tissue distribution will be iteratively updated during the image reconstruction process.

-       Implement an optimization strategy to update the assumed tissue distribution in a defined manner until a good agreement between simulated data and ground truth has been found.

Literature

[1]

P. M. Meaney, S. D. Geimer, and K. D. Paulsen, “Two-step Inversion with a Logarithmic Transformation for Microwave Breast Imaging,” Med Phys, vol. 44, no. 8, pp. 4239–4251, Aug. 2017, doi: 10.1002/mp.12384.

Knowledge of Python or Matlab. A basic mathematical understanding of optimization problems is beneficial.