CT scan projection data artifact correction method and CT image reconstruction method

Abstract

The invention proposes a CT scan projection data artifact correction method and a CT image reconstruction method. The CT scan projection data artifact correction method comprises the following steps:respectively acquiring thin and thick layer projection data of the same object; then constructing a neural network model for eliminating at least one artifact; training the neural network model with the thick-layer projection data as the input and the corresponding thin-layer projection data as the output; and performing artifact correction on aliased projection data with artifacts by using the trained neural network. Image reconstruction is performed on the projection data after artifact correction to obtain an artifact-corrected image. The correction method can significantly reduce aliasingartifacts, and at the same time can improve the resolution of an original CT image, no other hardware update replacement is required, and thereby the cost can be significantly reduced.

Description

technical field [0001] The invention relates to the technical field of CT scan image processing, in particular to a CT scan projection data artifact correction method and a CT image reconstruction method. Background technique [0002] CT (computed tomography imaging system) uses X-rays to scan objects, data acquisition to obtain projection data, and process these projection data through tomographic reconstruction algorithm to obtain the object's section and three-dimensional density information, so as to achieve the purpose of non-destructive testing. The third generation CT data acquisition system such as figure 2 As shown, S is the ray source, and D is the detector. Adjusting the position of S can correspond to different view direction sampling, including: viewing angle (view) direction sampling, detector channel (channel) direction sampling and detector row (row) direction Sampling (multiple row CT system). Throughout the sampling process, some rules must be followed to...

AI Technical Summary

Problems solved by technology

But quarter detector offset double sampling relies on two sets of samples taken 180 degrees apart, so any patient motion in the sampling interval will negate the aliasing reduction effect

In addition, except for the central ray, the double sampling at other positions is only approximate, so the effect of reducing aliasing artifacts depends on the position; the femtofocus technology in the X and Z directions of the tube requires the support of advanced ray sources and higher cost

[0008] In order to solve the aliasing artifacts in the row direction, a method of thick layer reconstruction and Z-direction flying focus is proposed. Thick layer reconstruction can reduce the artifacts caused by aliasing in the row direction, but it will lose the image resolution in the Z direction.

Images