MRI Technology Research Based On Compressed Sensing

Magnetic resonance imaging (MRI) is an important tool for medicalimaging, the data acquisition process is very slow. CS applied to MRI can provide asignificant potential to reduce the scan time, bring health benefits for patients. In thispaper, the compressed sensing theory is applied to image reconstruction, in-depthstudy of existing sample matrix and the existing algorithms, conduct research in thefollowing areas for its slow rebuilding and reconstruction quality is not highdrawbacks:(1) Carried out a study about sample matrix, and proposed a random anglesampling scheme. This paper studies several commonly used sampling matrix ofcompressed sensing theory, the angle between adjacent sampling line is the same formirroring sampling matrix,but there is a high coherence between the sample matrixand image sparse matrix,so this paper propose random sampling around samplingline center and designed a random sampling interval angle,this matrix satisfy aGaussian random distribution. Then through simulation experiments to ensure thisimprovement program has a lower correlation between sample matrix and sparsematrix, and can be accurately reconstructed image.(2) Conducted a research of iterative threshold algorithm and improved it. Thispaper studies reconstruction of several commonly used algorithms of compressedsensing theory, the soft threshold function of iterative has a constant bias between theestimated signal and the measured signal, and when the absolute value of the measuredsignal is less than the threshold value,all the estimate information set to zero, causingmany details of estimated signal missing’s issues.So this paper proposed adaptivethreshold iterative algorithm, and through simulation experiments prove thisimprovement program to ensure image reconstruction with high quality. (3) Research and design a new three-dimensional sample matrix,then apply it tothree-dimensional magnetic resonance imaging. Currently compressed sensingapplications in magnetic resonance imaging is basically limited to two-dimensionalimages, the study of the compressed sensing used in the three-dimensional magneticresonance images is very few.This paper research and design a new three-dimensionalsample matrix, the three-dimensional characteristics of the sample matrix is to fullysample in the low-frequency part (low frequency part through a solid cylindricalsampling, and the size of the cylinder can be flexibly adjusted), the high-frequency partwith random sampling (high-frequency part’s sampling line parallel line the lowfrequency portion’s solid cylinder). Shorten the time to bring the benefits of reducingthe reconstruction of sampling points. And prove the validity of this three-dimensionalmatrix of the sample through simulation experiments.