Cone beam CT noise estimation and suppression method for neural network learning

Abstract

The invention provides a cone beam CT noise estimation and suppression method for neural network learning, and the method comprises the steps of constructing a simulated cone beam CT noise-containingprojection sample data set through the noise priori knowledge of a real cone beam CT projection domain, and achieving the recognition of cone beam CT noise features for neural network learning. According to the method, real cone beam CT noise estimation without manual intervention is completed in a self-adaptive mode through a network model trained by simulation data samples. The method provided by the invention is suitable for cone beam CT noise estimation and suppression of a measured object with any complex structure, the reliability and universality of the method are good, the interferenceand influence of cone beam CT noise artifacts on the image can be reduced to a great extent, and the quality of the cone beam CT image is obviously improved.

Description

technical field [0001] The invention relates to a method for estimating and suppressing noise of cone-beam CT based on neural network learning, and belongs to the fields of medical imaging and industrial non-destructive testing related to the application of cone-beam CT. Background technique [0002] Cone beam CT (Cone Beam Computed Tomography, CBCT), as an advanced medical imaging and industrial non-destructive testing technology, has fast scanning speed and high radiation utilization rate, and can accurately and intuitively scan two-dimensional or three-dimensional high-resolution tomographic data Display the internal structure of the detected object, and quantitatively provide the position and size of the internal structure of the object. [0003] The actual industrial ray detection process is coupled by multiple factors, including quantum noise, dark field noise, etc., resulting in the imaging information mixed with various artifacts and noise, the ray attenuation inform...

AI Technical Summary

Problems solved by technology

[0006] (1) Existing denoising techniques have features such as smoothing or even covering image object edges and textures, which affect the expression of original image information; at the same time, various models need to set different parameters, which increases the complexity of noise suppression

[0007] (2) Most methods mainly denoise natural images, and there is no research on noise suppression for cone beam CT imaging process of high-voltage and high-density industrial objects

[0008] To sum up, the existing technical methods have poor versatility, and are often subject to many restrictions in practical applications, which cannot meet the high-precision medical imaging requirements of cone beam CT and the high-efficiency industrial non-destructive testing requirements.

Images