Enhancing Visual Perception in Interactive Direct Volume Rendering of Medical Images

  • Author / Creator
    Sharifi, AmirAli
  • Accurate information is the foundation of correct clinical diagnoses. Physicians are increasingly relying on new devices and tools to improve the quality of information used in their decision making. Wrong, misleading, or hard to interpret data can prevent patients from receiving proper treatment, hence putting their lives in danger or negatively impacting their quality of life. Medical imaging is one of the most widely used techniques in clinical diagnosis. It enables physicians to view the underlying anatomy of the patients. The information attained using medical imaging provides a non-invasive approach to detection of abnormalities and defects. Many imaging modalities are widely used today such as Computed Tomography (CT), and Magnetic Resonance Imaging (MRI). These imaging techniques in particular provide physicians with volumetric data representing the segment of interest from the internal anatomy of the patient. An effective and accurate visualization of the volumetric data is one of the bottlenecks among many, that prevents physicians to take advantage of the full potential of the available information. Many visualization techniques such as creating a set of 2D slides, surface rendering, direct volume rendering (DVR), etc. has been used to make the volumetric information available for the consumption of the physicians. However, in the past 30 years, lack of proper algorithms as well as limitations of hardware have prevented us from using the full potential of methods such as DVR. DVR creates 3D rendered images based on the volumetric data. Human brain is naturally trained to understand the spatial arrangement of objects using our visual system. A 3D rendered image simulates our vision. In contrast, understanding of the spatial arrangement, sizes, and shapes of objects using 2D slices is far less intuitive and requires many years of training. Moreover, DVR uses all of the volumetric data, while methods such as surface rendering only use a fraction of the data while discarding most of it. Despite great advantages of DVR, its usage remained limited due to some of its shortcomings. DVR simulates the physical process by which the image of a 3D scene is formed on a 2D surface. However, it is not computationally viable to simulate all or even nearly enough rays of light to create naturally occurring phenomena such as depth of field (DoF). These natural phenomena are used by our brain alongside each other to provide us with visual cues to help us interpret the images we see. In the first phase of this research we have created an algorithm that enables any portion of the volume to be rendered with the desired degree of blur in real-time using DVR. This algorithm is then utilized to tackle the lack of DoF problem in DVR. We have taken advantage of parallel programming and a novel sampling scheme to minimize the rendering time of DoF-enabled DVR. In the second phase of this research we have used our blur algorithm to introduce two new synthetic visual cues to help physicians: Transfer Function Based Blurring (TFBB), and Focus-Based Color Coding (FBCC). The former technique augments transfer functions to control the degree of blur based on the material. This enables the physicians to create semantic depth of field, keeping the material of interest in focus while other blurred material provide context. FBCC provides a way to direct viewer's attention to the object of interest while clearly distinguishing objects of interest. The combination of depth of field, TFBB, and FBCC has also been examined and shown great potential to remove clutter from the scene. The third phase of this research puts our previous two contributions through a rigorous usability study, where physicians are presented by images rendered by our methods and their interpretation of these images as well as many parameters of their interaction are recorded and evaluated. Our usability study showed that our proposed methods provide physicians with an enhanced visualization that has the potential to significantly improve correct clinical decision making.

  • Subjects / Keywords
  • Graduation date
  • Type of Item
  • Degree
    Doctor of Philosophy
  • DOI
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.
  • Language
  • Institution
    University of Alberta
  • Degree level
  • Department
    • Department of Computing Science
  • Supervisor / co-supervisor and their department(s)
    • Boulanger, Pierre
  • Examining committee members and their departments
    • Famil Samavati, Faramarz (Computer Science, University of Calgary)
    • Boulanger, Pierre (Computing Science)
    • Noga, Michelle (Department of Radiology & Diagnostic Imaging)
    • Cheng, Irene (Computing Science)
    • Punithakumar, Kumaradevan (Department of Radiology & Diagnostic Imaging)