Ultrasound Computed Tomography Algorithms Based On Full Waveform Inversion

Ultrasound computed tomography(UCT)is a promising technique in medical imaging.Compared with traditional imaging methods,UCT is featured with no ionizing radiation,no need of contrast agent,large field of view,broad applicability,convenience,and relatively low cost.More importantly,UCT is capable of reconstructing quantitative parameter distributions(i.e.,sound speed distribution and acoustic attenuation distribution)of measured tissues.These parameters,when combined with appropriate prior knowledge,may help identify pathologically different tissues.Therefore,it is of great importance to investigate UCT and the corresponding algorithms in the field of medical imaging.In a variety of existing UCT algorithms,the full waveform inversion(FWI)method has gradually become a promising UCT implementation method due to the abilities of multi-parametric modeling and high-resolution imaging.The current FWI-based UCT algorithms have some limitations.(1)When reconstructing the media with a large change in the sound speed,an optimization process easily falls into a local minimum and a reconstructed sound speed image with low accuracy is obtained.(2)The reconstruction of sound speed is based on the assumption that density of media is homogeneous,ignoring the heterogeneity in media density and increasing the error of sound speed reconstruction.(3)It is complex to manufacture the circular ultrasonic transducer array(CUTA)for UCT data acquisition.In view of these limitations,based on the advantages of FWI-based methods in tomographic imaging,in-depth simulation research of UCT algorithms is conducted in this dissertation work.Specifically,following research aspects are investigated.(1)When the FWI-based UCT algorithm is used with single-frequency source,an optimization process easily falls into a local minimum to reconstruct a media with a large change in the sound speed(body tissues,such as knee joints,that have bone structure and soft tissue),and a reconstructed sound speed image with low accuracy is obtained.For this purpose,a UCT based on FWI with multi-frequency source is proposed,which fuses the effective information obtained from excitation sources of each frequency.The simulation results suggest that,when reconstructing the media with a large change in the sound speed,the proposed algorithm is capable of effectively preventing the optimization process to fall into a local minimum and a sound speed image with high accuracy is reconstructed.(2)The existing single-parametric UCT algorithms reconstruct sound speed by assuming that density of media is homogeneous,but this assumption is not satisfied in some scenarios.To improve the accuracy of sound speed reconstructed images,a FWI-based bi-parametric(sound speed and density)UCT algorithm is proposed,which introduces the parameter of heterogeneous media density to effectively improve the accuracy of sound speed reconstruction,and the analyzing method of media density is provided.The simulation results suggest that,compared with the single-parametric UCT algorithms,the bi-parametric UCT algorithm based on FWI is capable of effectively describing the characteristics of density distributions and improving the accuracy of sound speed reconstructed images.The adaptability of the algorithm is increased.(3)For the complex manufacturing process of the CUTA,a UCT method based on a square border ultrasonic transducer array(SUTA)is proposed.The simulation results suggest that,compared with the CUTA,sound speed images with similar accuracy can be reconstructed by the use of the SUTA.The structure of the SUTA is easy to implement.