Incorporation of Ultrasonic Prior Information for Improving Quantitative Microwave Imaging of Breast

Abstract

Structural information derived via ultrasound is utilized as prior information for quantitative microwave imaging. The structural information is extracted from ray-based ultrasound reconstructions using a K-means clustering algorithm and consists of three tissue regions (skin, adipose, and fibroglandular). Tissue-specific complex permittivity values are assigned to each region (i.e., the complex permittivity is homogeneous over each region). The regions are then incorporated as an inhomogeneous numerical background in a quantitative microwave imaging algorithm (contrast source inversion). This new approach is assessed using synthetic data obtained from several anthropomorphic breast models of various densities derived from magnetic resonance imaging breast images, all containing tumors. Imaging results are quantitatively evaluated based on the algorithm's ability to detect the tumors. The performance is tested with four different variations of the prior information: two variations of the structural information and two of the assigned permittivity values. The resulting ultrasound-microwave multimodality imaging approach substantially improves the fidelity and accuracy of the reconstructed internal structures relative to previous studies that used radar-based microwave techniques to extract the internal structural information. An improvement in the sensitivity of the imaging algorithm to malignant tissue is also observed.