Open AccessResearch on High Sparse Sampling CT Image Reconstruction Algorithm Based on Convolutional Neural Networks
Author(s) -
Jiqing Luo,
Luzhao Feng,
Zhuxi Liu,
Junan Zhu,
Lai Song,
Yao Wu
Publication year - 2021
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/769/4/042005
Subject(s) - artificial intelligence , projection (relational algebra) , computer science , iterative reconstruction , convolutional neural network , computer vision , sampling (signal processing) , process (computing) , distortion (music) , image quality , artifact (error) , image (mathematics) , algorithm , pattern recognition (psychology) , amplifier , computer network , filter (signal processing) , bandwidth (computing) , operating system
CT plays an important role in medical diagnosis and industrial nondestructive testing. How to reduce the radiation to patients in the process of CT scanning has become a hot spot. An effective way is to use sparse sampling projection data to reconstruct the CT image. However, the image obtained by using the traditional CT reconstruction algorithm to deal with the sparse projection data has serious image distortion and artifact. This paper presents a CT reconstruction method based on CNN. This method learns the mapping relationship between the sparse projection data and the complete projection data in the training database, and uses the learned result to process the sparse sampling projection data. FBP algorithm is used to get the high resolution CT image. This paper uses CNN to carry out an end-to-end learning to repair the projection sinusoidal image with missing angle and improve the quality of CT image reconstruction.