Research On Microwave Induced Thermoacoustic Tomography In Heterogeneous Media

Microwave Induced Thermoacoustic Tomography（MITAT）is a hybride method which has great potential for biomedical imaging.It contains multi-physics conversion pro-cesses:tissues are radiated by microwave pulses;due to different pathological conditions of the tissues which have a strong correlation with the absorption of microwave energy,tissues generate sound pressures after thermal expansion and produce different ultrasound signals.These ultrasound signals are collected by ultrasonic transducers and used for imaging.Since the dielectric properties and the acoustic properties of the targets to be reconstructed maybe non-uniform,imaging algorithms in complex media and some ex-ploratory MITAT experiments are researched in this thesis.The novelty contribution of this thesis is summarized as follows:1.Microwave induced thermoacoustic imaging algorithm based on the characteris-tics of targets distributionFirstly,the regularization parameter of the traditional compressed sensing model in MITAT is redefined according to the measured data set.The reweighted?₁minimization is used to enhance the sparsity of the target distribution result.Secondly,the sparse char-acteristics of the thermoacoustic signals in the time domain and the target distribution in the spatial domain are considered in the signal model.The target vector to be estimated is divided into blocks.?_2/1regularization instead of the traditional?₁regularization is uti-lized.The subsequent optimization is processed by alternating direction multiplier method for each block in parallel,which improves the efficiency of the algorithm and saves the imaging time while ensuring the imaging quality.2.Microwave induced thermoacoustic imaging algorithm based on the spatial im-pulse response of transducerThe properties of ultrasound transducer are not considered in traditional imaging al-gorithms which only model the transducer to a point.In this thesis,the geometry character-istic of transducer is integrated into the forward model.In this case,to reduce the amount of calculation,time shifting is performed to generate a dictionary quickly based on the geometric relationship between the transducer and the target.Considering the spatial im-pulse response and the aperture of transducer,model-based algorithm can reconstruct the initial sound pressure distribution more accurately.3.Microwave induced thermoacoustic imaging algorithm based on statistical recon-struction according to the correlation of thermoacoustic signalsA statistical reconstruction algorithm based on correlation matrix is proposed to solve limited view problem.The algorithm uses the sparse Bayesian learning framework to sta-tistically reconstruct the thermoacoustic signals,which can not only reflect the estimated target distribution but also obtain the estimated error.And because the statistical recon-struction algorithm can make better use of all the prior information,the algorithm con-siders the block distribution characteristics of the imaging target and the correlation of thermoacoustic signals,which improves the robustness of reconstruction while ensuring the imaging quality.4.Microwave induced thermoacoustic imaging algorithm based on speed of sound autofocus according to Gaussian mixture modelA reconstruction method based on speed of sound autofocus is proposed to reduce acoustic inhomogeneity of different soft tissues.In this method,the number of tissue types（clusters）can be estimated through a decision graph.To distinguish the boundaries of different tissues,a Gaussian Mixture Model（GMM）is fitted to the image data for soft clustering instead of traditional hard clustering.Through fixing the tissue centers which are characterized by corresponding data density peaks in the decision graph as the means of Gaussian parameters rather than choosing them randomly,adaptive and robust recon-struction performance can be guaranteed.After an iterative GMM optimization method,the speed of sound autofocus is achieved.Image reconstructed by using the updated SoS distribution is more accurate than the result based on the homogeneous assumption.Com-pared with the existing methods,the proposed method does not need extra experiment costs,and it shows good robustness with respect to hard assignment model error when the medium is relatively complex.5.A series of microwave thermoacoustic imaging exploratory experiments are car-ried out.Experiments based on different microwave pulse widths is conducted to research the effects of microwave pulse width on imaging.Besides,the experiments of MITAT based on pig livers are performed to explore the possibility of MITAT for liver detection.In summary,MITAT in complex media is researched from imaging algorithm to ex-periments.Accuracy,efficiency and robustness of MITAT algorithms in complex me-dia are mainly studied.Numerical simulation and experiments based on practical tissue demonstrate the effectiveness of the proposed algorithm in this thesis.It provides a re-search basis for the application of MITAT in clinical.