Microwave tomography

Doctoral thesis, 2006

Microwave tomography for non-destructive detection of internal dielectric properties has received increasing interest during the last decades. Several application areas are under research, including applications in geoscience, biomedical applications, safety and security etc. The goal with this work is development of an microwave imaging equipment for breast cancer detection. Potentially this is very beneficial due to the large contrast in dielectric properties between the tumour and the healthy breast tissue. Within the frames of this work a microwave tomography prototype has been constructed. Since the image reconstruction is generally requiring computationally intensive optimisation algorithms comparison with a more computationally efficient algorithm, based on a low contrast approximation of the target, has been made. The results show that the reconstructions are comparable up to 10% contrast. For biomedical applications this is not sufficient and thus the iterative optimisation algorithms have been selected as the preferred algorithm for the rest of the work and as a platform for further developments. In the work with the experimental prototype some limitations in the performance of the algorithm have been identified. One is the inability to image large and small objects simultaneously, using a single electromagnetic pulse in the measurement. This is a problem in breast cancer tomography where the entire breast mass should be imaged together with potential tumours. The problem has been resolved by performing a reconstruction in several steps using successively increased frequency of the illuminating pulses. Another question that is addressed is whether it is feasible to utilise a priori knowledge about the dielectric properties of the object to improve the accuracy of the reconstructions. A reconstruction algorithm has been developed that is capable of reconstructing objects with given dielectric values. The results show that smaller objects can be resolved with the same frequency content of the illuminating pulse compared to when the a priori data is not taken into account.