Research On Ultrasound Tompgraphy Reconstraction Approach

Ultrasound tomography which is non-invasive, nonradiative, real time and low cost, scans as computer tomography, and it has wide application prospect in early diagnosis of breast cancer. The research of this paper is carried on around the ultrasound tomography reconstruction approach.First, a simulation of ultrasound tomography is computed in this reseach. According to theory of ultrasound tomography, a model contains fat, soft tissue, and tumour, reflects the characteristics of sound speed and density of tissues is established. Simulation is carried on with the model and primary data processing is also proceeded.Then, diferent reconstruction algorithms including filtered backprojection, algebraic reconstruction techniques, and two reflection algorithms are analysed and compared with each other. Simulation data is used to implement these algorithms to verify the analysis results. And accelerated reconstruction algorithm with GPU is achieved to cover long computing time.At last, a nonsubsampled controurlet transform based image fusion algorithm is applied to transmission tomography image which reflects tissue's inner characters well and reflection tomography image which has abundant edge information. Reflection image which is segmented with k-means clustering algorithm and sound speed image are fused in low pass sub-band of nonsubsampled contourlet transformation with segmentation result. The fusion result is compared with results of several common fusion algorithms through multiple objective indexes, and the conclusion shows our approach is the best.