CBCT Perfusion Based On C-arm CBCT System

Traditional CT is a structural imaging,whose imaging method can only be used to diagnose hemorrhagic stroke and cannot diagnose ischemic stroke in advance.But Ischemic stroke accounts for eighty-five percent of stroke medical records.Perfusion CT is a kind of functional imaging and is now generally used for stroke diagnosis.Due to CT’s less tomographic images and the longer transportation time from the imaging group to the intervention group,in order to ensure the best treatment time for the patient,it has extremely high application value to migrate the perfusion function from traditional CT to C-arm CBCT.However,the perfusion function after migration faces two new problems,namely time sampling and time resolution.Both of them are caused by the slower scanning speed of the C-arm,which can lead to deviations in imaging results and calculation errors in perfusion parameters.In addition,because of the high dose of the dynamic csanning,and the noise and artifacts caused by the use of low dose imaging which will cause errors in the calculation of perfusion parameters,so it is necessary to perform post-processing operations on the image after reconstruction.This article introduces traditional methods to solve the problem of time sampling and time resolution,and then uses deep learning for image post-processing to ensure the quality of imaging which builds a complete imaging process.To deal with time sampling problem,this paper proposes the idea of rollback reconstruction that the existing projection data can be reused and the number of reconstruction samples can be doubled without additional scanning time.In order to solve the problem of time resolution,this paper introduces the idea of partial block back projection,which is effectively combined with the idea of rollback reconstruction.The combined method is called RBTFDK.RBTFDK transfers the rollback reconstruction from the complete projection data to the partial block back-projection data,increasing the number of samples for partial block reconstruction while retaining the effect of the partial block back-projection algorithm for the time resolution problem.This article refers to the scanning protocol of C-arm CBCT,uses perfusion CT data to simulate the imaging process of perfusion CBCT and uses this data for time verification.The experimental results show that the RBTFDK algorithm combines the advantages of the rollback reconstruction and partial block back-projection algorithm,which can calculate more accurate perfusion parameter results while ensuring the imaging quality.After removing the influence of the time sampling problem and the time resolution problem,based on the U-net network structure,this paper proposes the NCS-Unet network to post-process the image.The NCS-Unet network separates the high frequency and low frequency of the feature map obtained after the input image convolution.Because the high frequency part obtained by decomposition contains not only the edges of the image,but also noise information,the Sobel filter is used in the network to extract the edge of the input image to increase the edge information of the network.The experimental results show that the NCS-Unet network proposed in this paper has a better noise suppression effect and can retain more detailed information.