Efficient dictionary construction method for microwave induced thermoacoustic compressive sensing imaging

Abstract

Microwave induced thermoacoustic imaging based on compressed sensing (CS) has proved to be a promising imaging method, especially in limited-view problems. However, the construction of the thermoacoustic dictionary in current CS-based methods may not be feasible for the real-world application due to the long construction time and large memory consumption. To address the cumbersome dictionary construction problem in the CS-based method, we propose an efficient solution by simplifying the forward solution of the wave equation which enables a fast calculation of the thermoacoustic dictionary through a time-shifting operation. In addition, by combining the temporal sparsity of the thermoacoustic signal, the dictionary can be compressed, resulting in a significant memory saving. We validated the proposed dictionary construction method using CS-based reconstruction by imaging numerical, biological phantoms and ex-vivo breast tissue and compared the image quality with the state-of-the-art reconstruction methods. A good image quality, fast construction time, and economical memory consumption demonstrate the potential of the proposed dictionary construction method in CS-based reconstruction for the real-world application.