The Research Of Random Coincidence Correction Method For Three-Dimensional Data In PET

Positron tomography imaging is also called positron emission tomography(PET)can be realized based on specific radioactive tracers.It is sensitive and precise and has a wide range of applications and development in neuroscience,pharmacology,clinical medicine,and other fields.To obtain quantitative images in PET,the correction for the data is an important step prior to reconstruction.This research paper focuses on the data correction for randomly occurring coincidence in 3D mode.This paper introduces the PET system for research and the necessity of correction for randomly occurring coincidence.The common methods of obtaining random coincidence measurement include single photon counting rate and delay window applied in this paper.The random coincidence measurements were applied in correction and reconstruction in two ways: one is to set the random data as the objective function and realize correction through iterative solution;the other is to calibrate the data with delay window measurement for correction.In order to avoid noise brought by statistics random coincidence,the paper researched the correction algorithm-PDEM algorithm(Prompt Delay EM)based on cross Poisson model.This method calculated random coincidence accounts of all channels with cross Poisson model of coincidence data and delays coincidence data in iteration.It did not involve estimating the correction process and avoid adding variance and additional noise.In order to verify the correction effect,the simulation experiment was conducted with 256?256 Shepp-Logan brain map as the standard map and iterative error under noise level 10%,30% and 50% as evaluation standard for correction.The results showed that PDEM' s correction effect was relatively stable under different noise ratios.The IEC simulation data and human body's liver data were applied for three-dimensional correction and reconstruction to further verify PDEM algorithm's correction effect of random noise in three-dimensional data through mean,standard deviation,and signal to noise ratio of the image to compare the correction effect of the PDEM.However,joint Poisson distribution model's correction effect of human body's three-dimensional data was imperfect because the other corrections were not implemented as well as the astringency of joint Poisson algorithm was limited during the process of three-dimensional reconstruction.Currently,the most widely used method is the delayed coincidence data for statistics random coincidence.Because delayed coincidence data has statistics variance and other noise,the single-plane Casey SP-C was researched which can reduce noise in delayed coincidence data.After getting the estimate of random coincidence data,subtraction correction and division correction combined with ML-EM algorithm were separately researched.In order to verify the validity of SP-C in both subtraction and division correction,the experiment is divided into three parts: first,the simulation experiments to verify the accuracy of correction subtraction division;then phantom data and human data analysis,in which correction methods were subtraction,division,subtraction correction method combined with SP-C,division correction combined with SP-C to prove the effectiveness of SP-C;and finally,the calibration experiments to compare the methods of SP-C and joint Poisson distribution model.