3D reconstruction for percutaneous interventions

Visualizing coronary arteries and stents is important during percutaneous coronary diagnosis and intervention. Nowadays interventional cardiologists typically use two-dimensional (2D) projection images and barely any three-dimensional (3D) images based on rotational X-ray angiography. The research in this thesis focusses on localizing, reconstructing and visualization of both contrast filled coronary arteries and stents in three dimensions. An important aspect of this is automated generation of the 3D data and visualization, such that the interventional cardiologist or assisting nurse is not bother with additional time consuming interaction with the system that is also prone to inducing errors. Background The main causes of death in the industrialized world remain cardiovascular disease (CVD) and cancer. Coronary artery disease (CAD), a subgroup of CVD, causes approximately 17.5% of all deaths in the US. During diagnosis and treatment of CAD it is important to visualize the coronaries in 3D and make measurements in 3D, because depth information is lost in a single 2D projection image. If a narrowing of a coronary artery needs treatment, the 3D information is important for accurate stent positioning, stent selection, and stent deployment. Current tools such as coronary modelling, quantitative coronary analysis (QCA) or intravascular ultrasound (IVUS) add additional user interaction or costs to a percutaneous coronary intervention. Stent reconstruction and visualization The first part of this thesis deals with motion-compensated 3D reconstruction of implantable devices, such as coronary stents and other cardiovascular devices. Chapter 2 gives an overview of various methods to enhance and visualize stents in both 2D projection images and 3D reconstructed volumes. Balloon-markers of the stent delivery system are used to recognize the stent region in the projection images. Using 2D motion compensation and temporal integration of multiple 2D images it is possible to enhance the stent in those 2D images, resulting in a better signal-to-noise ratio. This is important, because stents are often not symmetrically deployed, or not well deployed to the vessel wall. The deployment can often be improved by an additional balloon inflation inside the stent. Similar 2D motion compensation techniques can be used for rotational X-ray angiography in order to generate motion compensated 3D reconstructions. These 3D reconstructions can be viewed from any direction with limited user interaction. It is even possible to automatically generate virtual cross-sections of the deployed stent to check the symmetry and deployment of the stent. Chapter 3 describes the automatic detection of balloon-markers or other marker-like structures of stents can be automatically detected to enable 3D reconstruction during an intervention. An automatic visualization method is also described that shows cross-sections of a stent. In these images the deployed shape of the stent is easier to interpret, than by just looking at the outside appearance of a 3D reconstructed stent. Calcified plaques can also be identified in the reconstructions and cross-sectional images. Calcified plaques cannot always be pushed away during balloon-inflation and can therefore cause stent underdeployment. The described method is also successfully applied to other implantable devices. Chapter 4 describes a novel method for 3D motion compensation instead of 2D motion compensation. Especially rotational motion of a stent in the human body cannot be corrected for with using the 2D motion compensation method. By compensating the acquisition geometry for the actual motion of the device, instead of solely compensating the projection images, a real 3D motion compensation can be achieved. This method works well for rigid devices and needs a minimum of 3 markers for registration. Ex-vivo experiments show the difference of the 2D and 3D motion compensation method. The new method is also applied to an in-vivo closure device to close a septal defect. Chapter 5 concludes the first part of this thesis with some examples from clinical practice. These examples include a calcified vascular stent, Mitra-Clips for valves, a fractured coronary stent and a combination of coronary reconstruction and stent reconstruction. These real-life examples clearly show that this method is not only feasible for coronary stents, but also applicable to other implantable devices. It also shows clinically relevant aspects that can be visualized and quantified, such as an underdeployed stent or even a fractured stent. Coronary reconstruction and visualization The second part of this thesis deals with reconstruction of contrast filled coronaries. Chapter 6 gives an overview of coronary reconstruction, which is typically achieved by only reconstructing projection images that are in the same cardiac phase or with advanced motion compensation techniques. Visualization of the reconstructed coronaries and clinically relevant examples are described together with tools such as QCA, optimal working view determination, and road-mapping of for navigation of wires and devices, and stent deployment. Chapter 7 describes an intervention where rotational X-ray angiography and subsequent 3D reconstruction reveal and characterize a thrombus in the left coronary artery. In cases as this one the use of IVUS is potentially dangerous and rotational X-ray angiography has a clear benefit. The benefit of 3D reconstructions compared with 2D projection images is the ability to take a look inside the coronaries to get a clear view on the shape of the thrombus. Chapter 8 concludes the second part of this thesis with an assumed optimal acquisition scheme for percutaneous coronary interventions. An optimal trajectory is suitable for both diagnosis as well as 3D reconstruction and the trajectory should consists of the following: • contain as much as possible diagnostic information in the 2D images; • limited duration to limit the duration of the contrast injection; • and as much angular coverage along one axis as possible. The first results show that it is feasible to design such a trajectory on current C-arm systems and that diagnosis and 3D reconstruction, including additional tooling, are feasible. Discussion en conclusion Rotational X-ray angiography is not routinely used during percutaneous coronary interventions. The research in this thesis shows it is feasible to automatically generate 3D reconstructions of contrast filled coronary arteries and implantable devices such as coronary stents. Clinically relevant visualization and analysis techniques were developed to use these 3D reconstructions efficiently and intuitively during interventions. This adds new valuable tooling to the interventionalist performing percutaneous interventions, such as road-mapping. Performing more clinical studies and making this technology commercial availability are the next steps. This technology has the potential to reduce contrast agent usage, reduce procedure time, reduce X-ray dose, reduce costs, and, probably most important of all, increase patient outcome.