Energy spectrum CT image domain material identification method based on 3D full convolutional neural network

Abstract

The invention discloses an energy spectrum CT image domain material identification method based on a 3D full convolutional neural network. The method comprises the following steps: S1, acquiring energy spectrum CT images of different energy sections, wherein the energy spectrum CT images comprise to-be-identified material; S2, converting the energy spectrum CT images of the to-be-identified material in the different energy sections into three-dimensional image data; S3, inputting the three-dimensional image data into a 3D full convolutional neural network material identification model to obtain a material identification result of the energy spectrum CT image of each energy section; and S4, performing image fusion on the material identification result of the energy spectrum CT image of eachenergy section to obtain a comprehensive material identification image. According to the invention, a 3D full convolutional neural network is used to learn CT image information between different energy sections and features between different layers in a convolution process. An output image is subjected to a feature fusion method to obtain an energy spectrum CT image integrating the clear featuresof each energy section, so that an accurate and objective material identification result can be obtained.

Description

technical field [0001] The invention relates to the field of image recognition, in particular to a material recognition method in energy spectrum CT image domain based on 3D full convolutional neural network. Background technique [0002] X-CT (X-ray Computed Tomography), that is, X-ray computerized tomography or tomography imaging technology, is a nuclear imaging technology that can non-destructively detect the internal structure and material composition of the measured object by using the physical properties of X-rays. , has been widely used in many fields. The X-ray energy spectrum CT is a kind of computed tomography imaging with the help of X-ray energy-resolving photon counting detectors, using the information carried by the incident X-rays of different energies and the transmitted X-rays after the action of the detected object. technology. [0003] Spectral CT uses photon counting detectors, which can collect spectral information of multiple energy segments in one sc...

AI Technical Summary

Problems solved by technology

However, when photon counting detectors detect X-ray photons, they are affected by Compton scattering, charge sharing, pulse stacking effect, and photon noise, which lead to a decrease in the signal-to-noise ratio of CT reconstructed images and the accuracy of material decomposition.

It is difficult to distinguish material properties with similar densities in both the projection domain and the image domain, making it impossible to accurately identify a variety of soft tissue materials

[0004] At present, the analysis of CT images is mainly through the manual analysis of a large number of collected CT images by doctors. As we all know, analyzing a large number of CT images is a heavy work, and there is a certain degree of subjectivity, which is easy to cause misdiagnosis

Images