Improved compressive sensing of microwave induced thermo-acoustic tomography for breast tumor detection

Microwave induced thermo-acoustic tomography is a hybrid method that takes advantages of high contrast and high resolution. The method has attracted wide attention because of its potential ability to detect early breast cancer. To decrease the detection period, compressive sensing was used to reduce the intensive spatial sampling and temporal sampling. In our previous work, a method termed CS-MITAT was presented. Aiming to further enhance the image contrast of MITAT and improve computational efficiency some improvements are made in this paper based on CS-MITAT. The regularization parameter of the objective function, which has significant effects on reconstructed image quality, is afresh selected according to the corresponding data in MITAT based on generalized cross-validation (GCV). The weight matrix is also added to l
₁
minimization to enhance sparsity. The values of weights can be to neutralize the influence of signal magnitude on l
₁
penalty function. Due to its fast convergence, alternating direction method of multipliers (ADMM) is employed to solve convex unconstrained optimization problem. Compared with gradient projection for sparse reconstruction (GPSR) in previous work, simulations and experiments of real breast tumor demonstrate the contrast of reconstructed images is enhanced by the improved method and the computing time is also decreased.