Reconstruction of microwave tomography for breast cancer detection

Abstract

Abstract Breast cancer is the most prevalent cancer worldwide, affecting mostly women but men as well. The survival rates have increased dramatically over the last 50 years, primarily due to early detection protocols and enhanced treatment options. However, today’s screening techniques consisting mainly of x-ray mammography is both uncomfortable and contains harmful radiation. Depending on the breast type, it is also common with both false positives and fully omitted small tumours. Reconstruction using microwave tomography is a promising alternative, but is computationally heavy requiring tens of hours to reconstruct a single image and expensive hardware accurately collect the data. In this thesis, an approach using a discrete dipole approximation has been evaluated and compared to results from electromagnetic simulations in COMSOL Multiphysics as well as results from a finite-difference time-domain method. The suggested dipole method for reconstruction is a continuation of work conducted at Chalmers in collaboration with Dartmouth. The results are positive, indicating good detection for small objects, especially those with higher dielectric properties than the background material, which are exactly those objects omitted with traditional screening. The reconstruction of these objects are also good. However, for larger objects, especially those with lower dielectric properties than the background, and for more complex geometries, more evaluations needs to be performed.