Research On Low-dose CT Image Noise Reduction Based On Non-local Prior Knowledge Of Group Images

Although the low-dose CT(LDCT)technique can reduce the X-ray radiation damage to patients,CT imaging will have serious artifact noise interference and cannot meet the clinical diagnosis requirements.The individual patient’ previous normal-dose CT(NDCT)image assisted strategies can perform well in noise reduction of LDCT image.However,the requirement of repeated scanning of the same patient limits their practical application.In addition,as misalignments usually occur among the image series due to the involuntary body movement while scanning,there is a risk of the false priori structures may be introduced into the LDCT image.The clinically available population patients’ NDCT images(population images)contain valuable priori features such as textures and structures.If an imaging scheme can be used to extract priori features from population images and apply them to LDCT imaging,it can greatly improve the quality of LDCT imaging.This paper proposes the scheme of LDCT image noise suppression based on non-local priori information with population images.This scheme extracts non-local priori knowledge in the population images and uses it to adaptively normalize the target LDCT image,to achieve the purpose of improve the quality of low-dose X-ray CT imaging.The main works are summarized as follows:(1)The new scheme of low-dose CT image noise suppression based on non-local priori information with population images is proposed.The proposed scheme,including the methods of constructing population images database,priori sample search,non-local priori knowledge extraction and adaptive regularization of interest region of target image is studied systematically.(2)The algorithm of low-dose CT image noise suppression based on adaptive priori grayscale feature matching and non-local priori redundant information extraction from the population images is proposed(PFM-NLM).This algorithm constructs the offline feature database of population images,extracts the gray value information of the priori images to characterize the texture features of the image,and performs similar priori samples online search and non-local mean techniques.Furthermore,in order to efficiently suppress the noise in the LDCT image,the adaptive priori information mining and regularization processing are carried out for the local interest region of the target image.The performance of PFM-NLM algorithm was validated by CT images of patient with lung cancer.The results indicated that the proposed algorithm is superior to traditional algorithms in noise suppression and texture preservation,and can effectively avoid introducing false priori structures.(3)The algorithm of low-dose CT image noise suppression based on adaptive priori texturefeature matching and non-local priori redundant information extraction from the population images is proposed(PMFM-NLM).Based on the PFM-NLM algorithm,PMFM-NLM algorithm further extracts the five types of texture features(energy,entropy,contrast,correlation and inverse moment)calculated by the Gray-Level Co-Occurrence Matrix(GLCM)to represent the texture features of the image,and combines the voxel gray value feature of the image block to constructs the offline feature database of population images.This algorithm enhances the ability of the priori sample information expression and improves the search accuracy of similar samples.Finally,the non-local priori knowledge adaptive mining and regularization processing of the local region of interest of the target image is used to achieve low-dose CT image noise suppression.The performance of PMFM-NLM algorithm was validated by CT images of patient with lung cancer.The results show that the proposed algorithm can improve the quality of low-dose CT image compared with the traditional algorithms and PFM-NLM algorithm.