Microwave Breast Imaging with Prior Ultrasound Information

Abstract

Aiming at early detection of tumors, microwave breast imaging is investigated with a priori information on tissue boundaries yielded from ultrasound reflection data. A regularization term is to incorporate the information that two neighboring pixels should exhibit similar dielectric properties when not on a boundary while a jump would be allowed otherwise. This regularization is enforced in the distorted Born iterative method and in the contrast source inversion method. Comprehensive numerical experiments are carried out, involving a simple synthetic model and on two anatomically-realistic MRI-derived numerical breast phantoms. Imaging quality appears greatly improved with this regularization when tissue boundaries are indeed provided. Improvement is observed also when real parts and imaginary parts are retrieved in separate fashion.