Towards virtual electrical breast biopsy: space-frequency MUSIC for trans-admittance data

Breast cancer diagnosis may be improved by electrical immittance measurements. We have developed a novel method, space-frequency MUItiple Signal Classification (MUSIC), to determine three-dimensional positions and electrical parameters of focal lesions from multifrequency trans-admittance data recorded with a planar electrode array. A homogeneous infinite volume conductor containing focal inhomogeneities proved to be a useful patient-independent model for the breast containing focal lesions. Lesions polarized through the externally applied electric field are considered as distributions of aligned dipoles. Independence of the lesions´ shape and size is achieved by a multipole expansion of such a dipole distribution. Thus, lesions are described by point-like multipoles. Their admittance contributions are given by a sum over products of multipole-specific source-sensor transfer functions, called lead fields, multiplied by their moments. Lesion localization corresponds to multipole search, and uses orthonormalized lead fields for comparison with a signal subspace from a singular value analysis of a space-frequency data matrix. At the locations found, the moments´ frequency behavior is calculated which is assumed to be tissue-specific due to their dependence on conductivities. Results from clinical data show that space-frequency MUSIC successfully localizes lesions. Tissue differentiation might be possible, especially when the frequency range of the measurement system will be increased.