| 主权项 |
1. A method for reconstructing a PET image using GPU parallel computing, the method comprising the following steps:
1) sampling particles of each voxel using a linear sampling method according to a probability range of a voxel value, and obtaining an intensity value of each particle of the voxel; 2) calculating a prior intensity value of the voxel using an FBP method according to coincidence counting vectors acquired from sampling, and calculating a weight value of each particle of the voxel corresponding to the intensity value of each particle using GPU parallel computing and estimation algorithm of the weight value of the particle according to the prior intensity value of the voxel and the intensity value of the particle; 3) resampling the intensity value and the weight value of the particle using GPU parallel computing, and obtaining the resampled intensity value of the particle and the resampled weight value of the particle; 4) repeating step 2), using the resampled intensity value of the particle as the intensity value of the particle in step 2), repeating step 2) and step 3) until the resampled intensity value of the particle is converged to a certain value, and defining the resampled intensity value of the particle and the corresponding resampled weight value thereof as a true intensity value of the particle and a true weight value of the particle; and 5) calculating the voxel value using GPU parallel computing according to the true intensity value of the particle and the true weight value of the particle. |
