Microwave Induced Thermal Acoustic Tomography Based On Cluster Structure Prior In Breast Cancer Detection

Microwave induced thermal acoustic tomography(MITAT)is a new technique for breast cancer detection.This method combines the advantages of high contrast of microwave imaging(MI)and high resolution of ultrasonic imaging(UI),which shows a good application prospect in early breast cancer detection.Microwave induced thermal acoustic(MITA)signal is a kind of block sparse signal.Besides the sparsity of the signal,it also has the characteristics of cluster structure.Making full use of this structure feature is conducive to the establishment of a more accurate signal model,which can improve the reconstruction accuracy of the algorithm.Breast tumors are distinguished by images recovered from the algorithm,and the image quality is closely related to the imaging algorithm.Therefore,the research of MITAT algorithms based on prior cluster structure is of great significance for early detection of breast cancer.This paper mainly studies the MITAT algorithms based on the cluster structure prior in breast cancer detection.The main research contents are as follows:1.Compressive sensing(CS)is applied to MITAT to reduce data acquisition time.Specifically,gradient projection for sparse reconstruction(GPSR),Bayesian compressive sensing(BCS)and Total-variation compressive sensing(TVCS)methods are used to reconstruct tumor images.The simulation results show that TVCS can recover the massive tumor information more accurately than the other two methods.Comparing the images reconstructed by MI and MITAT under the same conditions,it is confirmed that MITAT has higher resolution than MI.2.MITAT algorithms based on Bayesian inference is proposed.These algorithms consider the cluster sparse characteristics of MITA signals,which non-zero coefficients of MITA signals are likely to accumulate into blocks.In the Bayesian framework,the sparsity and cluster prior of MITA signals are modeled separately.The Markov Chain Monte Carlo(MCMC)sampling and the variational Bayesian(VB)are used to implement Bayesian inference,and then two non-parametric algorithms are obtained to restore the original image of the tumor.Experimental results show that compared with MCMC sampling,the algorithm based on VB inference can achieve more accurate image reconstruction with shorter running time.3.This paper proposes a MITAT algorithm which combines generalized approximate message passing(GAMP)and VB.Rather than relying on the sparsity assumption,this paper introduces the hierarchical Truncated Gaussian Mixture(TGM)prior model and the cluster prior model to model the non-negative binary property and clustering structure characteristics of MITA signal,respectively.In order to improve the reconstruction efficiency of the algorithm,a generalized approximate message passing(GAMP)algorithm is embedded in the VB framework,which forms a method called Clu SSGAMP-VB.This method can realize the rapid recovery of images.Simulation results show that the proposed method not only further improves the reconstruction accuracy and robustness of the VB algorithm,but also greatly shortens the running time and improves the reconstruction efficiency.