Performance analysis for Bayesian microwave imaging in decision aided breast tumor diagnosis
In this paper the Markov Random Field is used to model the breast permittivity cross section as a propagating medium, and incorporate it into the forward Electromagnetic (EM) propagation to predict the random field of the EM measurements at a received array of sensors. Given these EM field measurements, Bayesian approaches are then developed to compute the likelihood ratio for tumor detection and the a posteriori probability display of tumor localization. Quantitative performance evaluations using simulations demonstrate the advantage of using the Bayesian approach to directly process the measurement data as compared to using the Bayesian or threshold approaches to detect and localize the tumor based on the reconstructed permittivity image.
