Realistic Microwave Breast Models Through T1-Weighted 3-D MRI Data

Abstract

In this paper we present an effective method for developing realistic numerical three-dimensional (3-D) microwave breast models of different shape, size, and tissue density. These models are especially convenient for microwave breast cancer imaging applications and numerical analysis of human breast-microwave interactions. As in the recent studies on this area, anatomical information of the breast tissue is collected from T1-weighted 3-D MRI data of different patients' in prone position. The method presented in this paper offers significant improvements including efficient noise reduction and tissue segmentation, nonlinear mapping of electromagnetic properties, realistically asymmetric phantom shape, and a realistic classification of breast phantoms. Our method contains a five-step approach where each MRI voxel is classified and mapped to the appropriate dielectric properties. In the first step, the MRI data are denoised by estimating and removing the bias field from each slice, after which the voxels are segmented into two main tissues as fibro-glandular and adipose. Using the distribution of the voxel intensities in MRI histogram, two nonlinear mapping functions are generated for dielectric permittivity and conductivity profiles, which allow each MRI voxel to map to its proper dielectric properties. Obtained dielectric profiles are then converted into 3-D numerical breast phantoms using several image processing techniques, including morphologic operations, filtering. Resultant phantoms are classified according to their adipose content, which is a critical parameter that affects penetration depth during microwave breast imaging.

Development of MRI‐based axillary numerical models and estimation of axillary lymph node dielectric properties for microwave imaging

“…But this is not always possible without compromising the segmentation of other tissues. Previous studies of breast or torso segmentation [25][26][27][28] have not included ALNs. Other studies [29][30][31] have addressed ALN segmentation but surrounding tissues were not segmented.…”

Development and application of human breast phantoms in microwave diagnostic and therapeutic technologies

“…Methods for deriving anatomically realistic numerical breast phantoms (e.g. [3]- [5]) employ the same basic overarching strategy. A spatial map of normal tissue types throughout the breast volume is obtained empirically from magnetic resonance imaging (MRI) of breasts of healthy women.…”

“…A two-component Gaussian mixture model was used to segment the voxel intensities and generate the linear mappings of each level. In [5] voxels were segmented into three levels, and nonlinear mappings between voxel intensity and dielectric properties were used to assign the properties. Figure 1 shows representative numerical breast phantoms from the University of Wisconsin online breast phantom repository [6].…”

Classification of rat mammary carcinoma with large scale in vivo microwave measurements

“…Inherent dielectric property discrepancy between the malignant breast tumors and healthy breast tissues at microwave frequencies 1 , 2 has fueled the research in development of microwave imaging and hyperthermia applications 3 , 4 , while potential other use of microwaves, such as microwave surgical margin detection and microwave biopsy, remains overlooked. Currently, both diagnostics applications are based on pathology and performed by specialists.…”