Heterogeneous platform-based CT image three-dimensional reconstruction acceleration method and device thereof

Abstract

The invention relates to a heterogeneous platform-based CT image three-dimensional reconstruction acceleration method and a device thereof. The heterogeneous platform comprises a host and heterogeneous OpenCL computation equipment. The method comprises the following steps of: carrying out computation particle decomposition on an FDK reconstruction algorithm and analyzing parallel computation flows of each computation particle; and carrying out acceleration optimization on each computation particle through the host and the heterogeneous OpenCL computation equipment in the heterogeneous platform. According to the method, the parallelism of a CT reconstruction algorithm is deeply mined, a GPU+FPGA heterogeneous calculation mode is adopted, and a computation system is formed by computation units which use different types of instruction sets and system architectures, so that the algorithm is matched with a heterogeneous structure system to the greatest extent, and the performance of different acceleration components is fully utilized; and meanwhile, storage and communication schemes suitable for efficient operation of the reconstruction algorithm are designed, the system supports PCI-E / Ethernet interconnection and supports a plurality of processing plates to realize efficient parallel processing of a plurality of processors through an interconnection bus, so that a synchronous or asynchronous cooperative processing mechanism is realized, and the reconstruction speed is improved under the premise of decreasing the loss precision as much as possible.

Description

technical field [0001] The invention belongs to the technical field of X-ray CT, and in particular relates to a heterogeneous platform-based acceleration method and device for three-dimensional reconstruction of CT images. Background technique [0002] X-ray computed tomography (Computed Tomography, CT) is a technology to reverse the attenuation distribution of an object from the X-ray projection, covering nuclear physics, mathematics, computers, precision instruments and other disciplines. Because CT can obtain high-precision three-dimensional structural information inside objects under non-contact and non-destructive conditions, since Hounsfield successfully developed the first CT, CT has been widely used in nondestructive testing, medical diagnosis, material analysis and other fields. [0003] In practical applications, the computing resources and storage resources required for high-resolution 3D reconstruction of cone beam CT are very large. With the increase of the reco...

AI Technical Summary

Problems solved by technology

[0003] In practical applications, the computing resources and storage resources required for high-resolution 3D reconstruction of cone beam CT are very large. With the increase of the reconstruction scale, the storage requirements and calculation amount of the reconstruction increase rapidly. In many cases, it is difficult to meet the To meet the needs of practical applications, take the common back-projection calculation in the reconstruction algorithm as an example: if the size of each dimension of the three-dimensional image to be reconstructed is N, the computational complexity of the corresponding back-projection will be as high as O(N4), and the reconstruction of a resolution of For a 10243 three-dimensional image, the calculation will cycle 1099.5 billion times. It is very time-consuming to complete such a large amount of calculation on an ordinary PC, and it is difficult to meet the requirements of practical applications.

Images