X-ray computed tomography energy spectrum estimation method based on convolutional neural network

Abstract

The invention provides a convolutional neural network-based X-ray computed tomography energy spectrum estimation method. The method comprises: step 1, designing a convolutional neural network for describing the coupling relationship between the CT image and the photon distribution probability in the energy spectrum; step 2, using the constructed training data set to train the convolutional neural network to obtain The trained convolutional neural network; step 3, using the trained convolutional neural network to estimate the power spectrum of the CT image to obtain CT image power spectrum information. This method mainly estimates X-ray energy spectrum information through the features of CT reconstruction images from the perspective of the image domain. It does not require additional hardware or a dedicated workflow, and can be directly applied to existing CT scanning systems, which is conducive to improving efficiency.

Description

technical field [0001] The invention relates to the technical field of CT imaging, in particular to a convolutional neural network-based X-ray computed tomography energy spectrum estimation method. 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. This technology uses the X-ray transmission information of the object at different angles, and obtains the attenuation characteristic distribution of the object through the image reconstruction algorithm, so as to realize the three-dimensional internal structure of the object without loss. Since Hounsfield successfully developed the first CT, CT has been widely used in non-destructive testing, medical diagnosis, material analysis and other fields. Among them, the X-rays generated by the tube in conventional CT ha...

AI Technical Summary

Problems solved by technology

The model-based method mainly uses empirical formulas or semi-empirical physical models to generate energy spectra, but for specific different ray sources and different energy intervals, the adaptability of the model method is limited

Measurement-based methods mainly estimate the used ray energy spectrum by indirectly measuring phantom attenuation information, which is relatively simple and has been implemented in a wide range of X-ray CT systems, but it requires auxiliary hardware, such as wedge models, resulting in the entire CT System scans require the addition of dedicated workflows

[0004] In 2008, Zhang Peng and others attributed the indirect measurement of energy spectrum to an optimization problem with regularization terms, but the regularization parameters in the model need prior information, otherwise the estimation results will be large

However, this method requires additional scanning of the auxiliary phantom, which makes the process of energy spectrum estimation more complicated, which is not conducive to direct application to existing imaging systems

Images