Mitigating acoustic heterogeneous effects in microwave-induced breast thermoacoustic tomography using multi-physical K-means clustering

Microwave-induced thermoacoustic tomography shows great potential for early-stage breast tumor detection, but imaging quality usually suffers due to acoustic heterogeneity of breast tissue. To mitigate this problem, conventional methods estimate the distribution of speeds of sound but at a heavy cost of system complexity or computation burden. We propose an imaging reconstruction method that incorporates dielectric and acoustic properties of tissues as a-priori information and reformulates the velocity estimation problem as a data clustering problem. The proposed method is validated by imaging anatomically realistic numerical breast phantoms and real biological tissues. Both simulation and experimental results demonstrate that the proposed method is robust and significantly improves image fidelity with less computational burden than conventional methods. These results make our approach a promising candidate for clinical breast cancer detection.