An unregistered low-dose CT denoising method based on an adversarial generative network and a computer

Abstract

The invention belongs to the technical field of medical image processing, and discloses an unregistered low-dose CT denoising method based on an adversarial generative network and a computer. The method comprises; acquiring LDCT data and NDCT data; Analyzing the data, and dividing the data into a training data set and a test data set in proportion; carrying out Programming in TensorFlow to realizea network framework; reading Data in and preprocessed, and adjusting the sizes of the images to be the same; Inputting LDCT into the two generators respectively to obtain a noise result and a noise suppression result respectively, and adding the noise result and the noise suppression result to obtain a false LDCT; Using the two discriminators to respectively discriminate the result after noise suppression and the false LDCT; Calculating loss functions of the two generators and the two discriminators through the generation result and the discrimination result; Optimizing the network through anoptimization algorithm to obtain a network with trained parameters; And testing on the test set to obtain an LDCT noise suppression result. The method can be used for the noise suppression problem ofunpaired data and the noise suppression problem of paired data.

Description

technical field [0001] The invention belongs to the technical field of medical image processing, and in particular relates to a denoising method and a computer for unregistered low-dose CT based on an adversarial generation network. Background technique [0002] Currently, the existing technology commonly used in the industry is this: X-ray computed tomography (CT) is one of the most important imaging modalities in modern hospitals and clinics. There is a potential radiation risk to the patient, as X-rays can cause genetic damage and induce cancer with a probability related to the radiation dose. Over the past decade, the dose of CT examinations has gradually decreased. Coronary CT angiography has decreased from about 12mSv in 2009 to 1.5mSv in 2014. Reducing radiation dose increases noise and artifacts in reconstructed images, compromising diagnostic information. Significant efforts have been made to design image reconstruction or image processing methods for low-dose CT...

AI Technical Summary

Problems solved by technology

[0004] (2) Prior art 1 and prior art 2 only focus on paired data, that is, the strict correspondence between pixels between LDCT and NDCT, while clinical data is unpaired data, that is, there is no corresponding relationship between pixels of LDCT and NDCT

[0005] (3) The third application of the existing technology is not effective in LDCT noise reduction, mainly because the details of the image generated by the network are not clear

This is reflected in the blurred boundaries of bone and soft tissue in the generated image, and the low peak signal-to-noise ratio and structural similarity compared with the original image, which affects the diagnostic results

Images