Virtual Heart and Vascular Endoscopy Using, Ultra-sound, CT, and MRI Data

Amir Sharifi (Supervisors: Dr. Boulanger and Dr. Noga)

There have been many advances in the field of medical imaging and radiology in recent years. New methods and algorithms as well as new generations of imaging devices are being introduced by scientists and companies. Modalities like MRI and CT can provide physicians with volumetric data representing the underlying anatomy and physiology of the patient. However, the use of this data is often limited to viewing slices of the volume in 2D. In this project, we plan to build a 3 dimensional (3D) and 4D (temporal) virtual endoscopy system to explore MRI and CT data from the heart and the blood vessels. The inability of 2D images to convey 3D spatial and temporal relationships limits experienced imaging specialists to understand the true nature of the dataset while increasing the chances of misinterpretation by clinical physicians.

In cardiac sciences, the goal of new imaging technologies is to render more accurate and useful information about the cardiovascular system to the clinicians, which will ultimately help them make more informed and effective clinical decisions. It is the hypothesis of this thesis that 3D and 4D visualizations can be used as superior sources of information for more precise clinical decision making. Currently the realization of spatial and temporal relations of arteries and the detection of abnormalities are solely based on the interventional cardiologists’ ability to imagine this complex network in 3D by only looking at 2D slices, which requires many years of extensive training. It is apparent that mistakes and missing information can occur and can be detrimental or even fatal for the patient.

Minimally invasive and non-invasive methods of diagnosis like tomographic imaging are generally preferred in medicine. Proper treatment of heart diseases like vascular rings and slings, atherosclerosis, aortic dissection, etc. is aided when physicians are facilitated with proper information like high contrast, high resolution, and sharp 3D images. There are methods like contrast agent injection that can be used for increasing the contrast and visibility of the cardiovascular system. High-resolution depends on the machine and on the reconstruction algorithm. Synchronization of the imaging device with diastolic phase in cardiac cycle ensures getting sharp images. Despite all these methods, we believe we are far from using the full potential of the available data.

In this project, we plan to build a 3D and 4D virtual heart and cardiovascular endoscopy system using temporal CT and MRI images. Endoscopy is a practice in medicine that uses tubes with cameras called endoscopes to examine the hollow organs in the body. It provides a very comprehensible visual of the internal tubes like digestive system. However, invasive methods like inserting a catheter into the heart and arteries entail more risk. Our solution to this problem is to build a virtual heart endoscopy system. This method has several advantages over current methods of examining the cardiovascular system. It can provide physicians with spatial and navigation information; it is non-invasive, which reduces the risk for the patient. It can generate views that are impossible to achieve when using an endoscope and techniques like color coding can be used to provide more information to the cardiologist.

This is a joint project between Department of Computing Science and Department of Radiology & Diagnostic Imaging at University of Alberta. My co-supervisor, Dr. Michelle Noga is an expert in pediatric and cardiac imaging. She is a professor at Department of Radiology and the director of Servier Virtual Cardiac Centre (SVCC). During the thesis, she will provides us with data, feedback, visual validation of correctness of the generated models, and consultation about the ways to make this system more useful for cardiologists and surgeons.

There are several challenges to overcome in this project. The first problem is the correct visual­ization of the data. There are two approaches mainly used for the visualization of 4D data. The first approach is surface rendering. In this approach, a polygonal model is created and rendered based on the segmentation of the arteries and the interpolation between different slices of the volumetric data. Getting the correct automatic or semi-automatic segmentation and finding the correct interpolation function are two of the main challenges. One of the ma­jor problems in this rendering mode is that, in many cases, physicians want to see the surrounding tissue, i.e. in case of a polyp. The other approach to visualization is to use volume rendering. Since there is no segmentation involved and no data is lost. Volume rendering is computationally more expensive than polygon rendering but more accurate for medical rendering. GPU and parallel programming can be used have been shown to work in real-time.

There has been some research on surface and volume rendering, in the area of virtual colonoscopy. However, the nature of the diseases physicians are looking for in arteries is completely different than the ones in intestines. As a result, visualization of atherosclerosis in coronary heart disease is a completely different challenge than visualization of polyps due to their differences in visibility, material, shape, size, and nature. There are still many unsolved problems in this and related fields like virtual CT gastroscopy which makes fiber-optic endoscopy still superior to the virtual one. This means that there are still many shortcomings to resolve before we solely rely on non-invasive diagnostic methods. In this project, we will explore both of these visualization methods in depth to find the best approach for the requirements of the interventional cardiologist.

The second but equally important challenge is the navigation problem. There has been some research done on virtual colonoscopy that can be used as a basic foundation for cardiovascular endoscopy. For example, the centre line algorithm has been used for generating a path to create a fly-through simulation inside the intestines. There are, however, many differences between virtual colonoscopy and cardiovascular endoscopy, which make the transition non-trivial. The cardiovas­cular system forms a huge interconnected network extended and intertwined in 3 dimensions which is unique for each person. To find proper ways of navigating through and land-marking arteries is one of the major challenges and is significantly more complex than the  single tube and relatively static shape of intestines.