Microwave imaging has been investigated as a method of non-invasively estimating tissue electrical properties especially the conductivity, which is highly temperature dependent, as a means of monitoring thermal therapy. The technique we have chosen utilizes an iterative Gauss-Newton approach to converge on the correct property distribution. A previous implementation utilizing the complex form (CF) of the electric fields along with a sub-optimal phantom experimental configuration resulted in imaging temperature accuracy of only 1.6°C. Applying the log-magnitude/phase form (LMPF) of the algorithm has resulted in imaging accuracy on the order of 0.3°C which is a significant advance for the area of treatment monitoring. The LMPF algorithm was originally introduced as a way to reconstruct images of large, high-contrast scatterers as is the case in breast imaging. However, recent analysis of the Jacobian matrices for the comparable implementations has shown that the reconstruction problem in the new formulation more closely resembles a linear task as is the case in x-ray computed tomography. The comparisons were performed by examining plots of the Jacobian matrix terms for fixed transmit and receive antennas which demonstrated higher sensitivity in the center of the imaging zone along with narrower paths of senstivity between the atnenna pair for the LMPF algorithm. Animal model experiments have also been performed to validate these capabilities in a more realistic setting. Finally, the overall computational efficiency has been significantly enhanced through the use of the adjoint image reconstruction approach. This enables us to reconstruct images in roughly one minute which is essential if the approach is to be used as a therapy feedback mechanism.
