A dual modality, DCE-MRI and x-ray, physical phantom for quantitative evaluation of breast imaging protocols

Abstract

The current clinical standard for breast cancer screening is mammography. However, this technique has a low sensitivity which results in missed cancers. Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) has recently emerged as a promising technique for breast cancer diagnosis and has been reported as being superior to mammography for screening of high-risk women and evaluation of extent of disease. At the same time, low and variable specificity has been documented in the literature as well as a rising number of mastectomies possibly due to the increasing use of DCE-MRI. In this study, we developed and characterized a dual-modality, x-ray and DCE-MRI, anthropomorphic breast phantom for the quantitative assessment of breast imaging protocols. X-ray properties of the phantom were quantitatively compared with patient data, including attenuation coefficients, which matched human values to within the measurement error, and tissue structure using spatial covariance matrices of image data, which were found to be similar in size to patient data. Simulations of the phantom scatter-to-primary ratio (SPR) were produced and experimentally validated then compared with published SPR predictions for homogeneous phantoms. SPR values were as high as 85% in some areas and were heavily influenced by the heterogeneous tissue structure. MRI properties of the phantom, T1 and T2 relaxation values and tissue structure, were also quantitatively compared with patient data and found to match within two error bars. Finally, a dynamic lesion that mimics lesion border shape and washout curve shape was included in the phantom. High spatial and temporal resolution x-ray measurements of the washout curve shape were performed to determine the true contrast agent concentration as a function of time. DCE-MRI phantom measurements using a clinical imaging protocol were compared against the x-ray truth measurements. MRI signal intensity curves were shown to be less specific to lesion type than the x-ray derived contrast agent concentration curves. This phantom allows, for the first time, for quantitative evaluation of and direct comparisons between x-ray and MRI breast imaging modalities in the context of lesion detection and characterization.