X Ray Key Technology Research For Image Guided Therapy In Vessel Intervention

X ray imaging is a image modality which obtains the internal structures of objects by detecting the different attenuation of X ray when it goes through different matters.X ray imaging can be used to obtain 3D anatomical structures of human body for disease diagnosis and intervention path planning,and can be used to obtain real time images of interested areas during intervention procedures.Therefore X ray imaging plays a significant role in image guided intervention therapies.In this paper,we choose two representative application of X ray imaging in vessel intervention therapies to study.We analyze the advantages and weakness of currently applied technologies,and propose our solutions for the challenges they are facing.With the introduction of flat panel detector technology,it is possible to obtain high quality X ray fluoroscopic images,which allows the C arm mounted CBCT 3D imaging to be widely applied.Currently one of the main challenges that CBCT technologies are facing is the relatively low calculation efficiency which fails to meet the clinical real time imaging requirement.In this paper,we propose a GPU based acceleration method for fast C-arm CBCT3 D image reconstruction.We apply an optimization of the 3D reconstruction algorithm based on its geometric features,and design a GPU based parallel acceleration plan to imporve calculation efficiency.Instead of the traditional linear image chain,we design a parallelly executed non-linear image chain system to apply the delay hiding technology to further lower3 d reconstruction delay.With the 3D reconstruction experiments using 2D projection image data sets simulated from digital phantom,and image data sets of actual imaging phantom acquired from C-arm imaging system,we proved that our proposed method can significantly increase 3D reconstruction efficiency without compromising reconstruction accuracy and image quality.X ray fluoroscopic imaging is a widely applied imaging modality to acquire anatomical structures of interested area during image guided interventions.Extracting the position and shape of the intervention tools such as guide wire,can enhance the intra operative imaging and better assist the intervention.Currently applied guide wire tracking and extraction methodshave main two shortcomings: manual intervention and low calculation efficiency.In this paper,we propose a GPU based automatic efficient approach to extract guide wire from fluoroscopic sequences.We apply a multi-scale filtering method to generate linear structure feature map from fluoroscopic image,and a gradient vector field method to converge the map to guide wire ridge points.We apply a GPU parallel acceleration method,and a Bayesian guide wire motion detection method to limit the field of interest to a smaller range to enhance calculation efficiency.We test our proposed method with two fluoroscopic sequence recording guide wire motions during cardinal vessel interventions,and show that compared with existing methods,our proposed method can automatically and more accurately extract guide wire structures,and the calculation efficiency is significantly improve to meet real time requirement.