Research On Photoacoustic Imaging In Complex Medium Based On Random Matrix Theory

Photoacoustic imaging(PAI)is a hybrid biomedical imaging technique,which combines the rich contrast of optical imaging and the high spatial resolution of ultrasonography,as well as the advantages of non-invasion and non-radiation.Among several modes of PAI,photoacoustic tomography(PAT)has been widely concerned as a more potential imaging technology in deep tissue.During the PAT imaging process,the optical absorbers in the biological tissue are irradiated by a laser pulse and then generate ultrasound as a result of the thermoelastic expansion.The ultrasound signal is picked up by the transducer array arranged outside the tissue and the images are reconstructed by applying a reconstruction algorithm.The classical image reconstruction algorithms used in PAT are usually based on the principle of coherent beamforming,which relies on an assumption of homogeneous media.However,acoustical inhomogeneity widely appears in various biological organs,and the difference in speed of sound will result in artifacts in the reconstructed images.In addition,the presence of bones,air cavities,or microbubbles may cause strong acoustic scattering and degrade image quality.Under this background,this paper focuses on the study of application of filtering algorithm in PAT to reconstruct images from the distorted and noise-like signals.Firstly,this thesis introduces the application of a filtering algorithm based on random matrix theory in acoustic imaging system.The single scattering signal and the multiple scattering signal in an echo model show the coherence and randomness respectively in the signal matrix,on the basis on which the corresponding filter can be designed to separate them.In this paper,the combination of the filtering algorithm with time reversal method and beamforming method is demonstrated respectively.The combination of the filtering algorithm with time reversal method can obtain better images in strong scattering media.Subsequently,the algorithm is theoretically extended from the linear transducer array case to the sector transduce array case,extending the application range of the filtering algorithm.The filtering algorithm has a good application prospect in the PAI system.However,the difference between the signal storage methods of the photoacoustic imaging system and the reflective ultrasonic imaging system has led to the fact that the filtering method based on the random matrix theory cannot be directly applied to the PAI system.In order to apply the filtering algorithm to PAT system,we construct a signal matrix from the signal vector detected by a linear transducer array to investigate the wide-band photoacoustic waves propagating through complex media.The direct and scattering wave coexists in the detected signals and we demonstrate that the intrinsic coherence of the direct wave can be revealed in the matrix while the scattering wave exhibits random characteristics in matrix form.A correlation filter is proposed based on this finding to reduce the scattering components in the detected signals.The accuracy and signal-to-noise ratio(SNR)of the imaging reconstructed by applying the time reversal method to the filtered matrix are improved as a result.For the proposed filter,the matrix interception in the process of matrix filtering makes some information discarded,resulting in image distortion and the reduction of imaging range.A correlation full-matrix filter is proposed to further improve the imaging quality and restore the imaging area.For the correlation filter proposed in the third chapter,the matrix interceptions during the matrix filtering process make the partial acoustic information not fully utilized,which results in the image distortion of the target and the reduction of field-of-view.The rotation and filtering operations are improved to make full use of all the useful information.In addition,a location factor is also proposed to compensate for the intensity unbalance in multi-targets imaging.The correlation full-matrix filtered imaging method is also applied to the annular transducer array.The circular array is divided into multiple sections and each sector array can be approximated as a linear array via phase and amplitude compensation.The image quality can be further improved by superimposing the imaging results of each section reconstructed by the correlation full-matrix filtered method.In the last chapter,the paper is summarized and the future development is prospected.