Tomographic reconstruction system

Abstract

A tomography system having a central processing unit, a system memory communicatively connected to the central processing unit, and a hardware acceleration unit communicatively connected to the central processing unit and the system memory, the hardware accelerator configured to perform at least a portion of an MBIR process on computer tomography data. The hardware accelerator unit may include one or more voxel evaluation modules which evaluate an updated value of a voxel given a voxel location in a reconstructed volume. By processing voxel data for voxels in a voxel neighborhood, processing time is reduces.

1 . A tomography system, comprising:
 a central processing unit;   a system memory communicatively connected to the central processing unit; and   a hardware acceleration unit communicatively connected to the central processing unit and the system memory, the hardware accelerator configured to perform at least a portion of an MBIR process on computer tomography data.   
     
     
         2 . The system according to  claim 1 , further comprising one or more voxel evaluation modules that evaluates an updated value of a voxel given a voxel location in a reconstructed volume. 
     
     
         3 . The system according to  claim 1 , further comprising an electronic display, the display operatively connected to the central processing unit, the electronic display configured to display a reconstructed image based on the computer tomography data. 
     
     
         4 . The system according to  claim 1 , further comprising a computer tomography scanner, wherein the computer tomography scanner is configured to irradiate a test object, measure resulting radiation, and provide measured data corresponding to the resulting radiation. 
     
     
         5 . The system according to  claim 1 , wherein hardware acceleration unit comprises a control unit which generates a pseudo-random sequence of voxel locations. 
     
     
         6 . The system of  claim 1 , wherein the hardware acceleration unit is configured to identify voxel data required to update a voxel and fetch the voxel data from the system memory. 
     
     
         7 . The system of  claim 2 , wherein the VEM contains VEM memory blocks internal to the VEM, which stores the data needed to compute the updated voxel value. 
     
     
         8 . The system of  claim 2 , wherein the VEM is configured to assess the data stored in the VEM memory blocks, re-use said data stored in the VEM memory blocks across multiple voxel evaluations, and partially fetch unavailable data from the system memory. 
     
     
         9 . The system of  claim 2 , wherein the VEM is configured to perform data transfer operations and data processing operations in parallel. 
     
     
         10 . The system of  claim 2  wherein the VEM is configured to perform data transfer operations and data processing operations in a pipelined manner. 
     
     
         11 . The system of  claim 2 , wherein the hardware accelerator unit fetches data required for a voxel from the system memory, while computations corresponding to a different voxel are in progress. 
     
     
         12 . The system of  claim 2 , wherein the hardware accelerator unit comprises a plurality of VEMs, the VEMs configured to update multiple voxels in parallel. 
     
     
         13 . The system of  claim 12 , wherein the sequence of voxels updated on a VEM is constrained to enhance data reuse within the accelerator. 
     
     
         14 . The system of  claim 12  in which at least one next voxel processed on a given VEM is constrained to lie within a common slice as the previous voxel processed on the VEM, thereby enabling the error sinogram memory to be reused. 
     
     
         15 . The system of  claim 12  in which voxels updated concurrently on multiple VEMs are constrained such that they share at least one entry of an A matrix of the tomography data. 
     
     
         16 . The system of  claim 15  where the said shared entry of the A matrix is fetched only once from the system memory and used by multiple VEMs. 
     
     
         17 . The system of  claim 12  in which adjacent voxels are updated on the same VEM, enabling neighborhood voxel data to be shared between the voxels. 
     
     
         18 . The system of  claim 12  where each VEM is configured to update a voxel neighborhood around a given voxel, the neighborhood comprising voxels adjacent to the given voxel.