Prior-based ROI reconstruction using weighted Hessian Schatten regularizations
作者单位：上海交通大学生物医学工程学院
会议名称：《第十七届中国体视学与图像分析学术会议》
会议日期：2022年
学科分类：0831[工学-生物医学工程（可授工学、理学、医学学位）] 08[工学] 080203[工学-机械设计及理论] 0802[工学-机械工程]
关 键 词：CT imaging ROI reconstruction weighted Hessian Schatten regularizations
摘 要：Computed tomography(CT) is widely applied in clinical diagnosis and its radiation dose has aroused public concern.Studies show that a higher cancer risk is closely related to more radiation exposure,which motivates many dose-saving strategies,e.***.automated tube current modulation,fewer views and regionof-interest(ROI) scan.ROI scan can be conducted in some specific medical applications,such as CT Angiography(CTA) and lung nodules follow-up examinations,wherein doctors pay more attention to the organs and structures than the surrounding tissues.Moreover,the shape and size of these structures are essential for diagnosis,since appearance changes often indicate diseases.In the ROI scan,the acquired data are truncated and noisy,resulting in an ill-posed reconstruction problem.Within the compressed sensing(CS) framework,the total variation(TV) penalty has shown excellent performance in edge preservation and noise suppression in ROI reconstruction.However,the TV penalizes the first-order derivatives and produces piecewise constant images,which probably conflicts with the characteristics of the underlying image and generates staircase effect.To avoid the over-smoothness induced by the TV penalty,we utilize Hessian Schatten regularizations to mitigate noise and artifacts,which penalizes the second-order derivatives and is perfect for reconstruction of piecewise linear images.Furthermore,in ROI reconstruction,weights based on the data deficiency are attached to the Hessian Schatten norm to compensate the inhomogeneity of back-projections.The structure-coupling method is adopted to update the ROI of the prior image from the reconstructed image iteratively to improve the quality of ROI reconstruction.In the experiment,a low-mAs,sparse-view,ROI scan was simulated,which greatly reduced the radiation dose.The prior images were obtained from the previous scans of the same patient.The results showed that the proposed method outperformed other TV-based methods,especially in structure preservation and tissue contrast.