Research On Algorithms For Reducing Artifacts In 4DCT Images Based On Principal Component Analysis

Due to the rapid advancement of computer technology and medical imaging technology,the accuracy requirements for cancer radiotherapy are also getting higher and higher.4D CT has been widely used to monitor patient-specific breathing movements to determine individual safety margins during radiotherapy.Due to the irregular breathing movement of patients during the acquisition of the 4D CT image,motion artifacts such as blurring,missing or irregular deformation of the scanned image are generated.Medical image registration is a very important part of medical image analysis,and its main application domain is the diagnosis and treatment of diseases.In the process of formulating patient's treatment plan,the rigid registration method is generally used to transfer the online image into the scanned image in order to correct the treatment plan.Spatial and temporal information are important to the structure of the image during four-dimensional medical image processing because the size,shape and location of the tumor are changing.For example,the respiratory movement of lungs and radiotherapy for patients can make the tumor morphology change,so it is necessary to consider the effect of time in the treatment progress.There is a variety of existing registration technologies,such as single-mode,dualmode,rigid body,non-rigid body,parameter-based and non-parameter-based,but some of these techniques ignore the nonlinear geometric part of the fluid and directly use the linear Euclidean space to calculate,and some of them have insufficient consideration of spatial and temporal information.A retrospective analysis is performed through the internal 4D CT database,and more than 50 treated patients are analyzed.It is found that nearly 70% images have artifacts.How to reduce and eliminate motion artifacts has become a popular topic in 4D CT reconstruction.In this study,DIR is used to track patients' anatomic and biological changes during radiotherapy.The main purpose of DIR is to find a transformation in two images with different phases.This transformation can reduce the difference between the two images,so as to minimize the deformed image and the target image by providing a voxel-to-voxel deformation matrix.The differences between the sets can be further applied to the deformation of the contours of the organs and the radiotherapy area and the calculation of the dose distribution.At present,DIR is widely used in radiotherapy,for example;a four-dimensional lung model can be used to track the movement and displacement of the lung over time.The foundation of this model is to collect a series of CT images of the lungs during continuous breathing.These CT images are registered by the corresponding DVF,so that each pixel of the lungs will have a specific mathematical function with time characteristics.Due to irregular movements of the patient such as breathing and swallowing,the 4D CT image will produce severe or mild motion artifacts.Motion artifacts have a serious impact on the quality of the image,and the degradation of image quality usually interferes with the doctor's diagnosis,leading to misdiagnosis or missed diagnosis,so eliminating and reducing motion artifacts is a very important link in medical image processing.In this study,a method of principal component analysis combined with a linear polynomial fitting model to process the displacement vector field obtained from the deformed image registration is proposed to reduce motion artifacts in the image.