Degree Type

Thesis

Date of Award

2019

Degree Name

Master of Science

Department

Electrical and Computer Engineering

Major

Human Computer Interaction; Computer Engineering

First Advisor

Eliot Winer

Abstract

The medical field has long benefited from advancements in diagnostic imaging technology. Medical images created through methods such as Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) are used by medical professionals to non-intrusively peer into the body to make decisions about surgeries. Over time, the viewing medium of medical images has evolved from X-ray film negatives to stereoscopic 3D displays, with each new development enhancing the viewer’s ability to discern detail or decreasing the time needed to produce and render a body scan. Though doctors and surgeons are trained to view medical images in 2D, some are choosing to view body scans in 3D through volume rendering. While traditional 2D displays can be used to display 3D data, a viewing method that incorporates depth would convey more information to the viewer. One device that has shown promise in medical image viewing applications is the Virtual Reality Head Mounted Display (VR HMD).

VR HMDs have recently increased in popularity, with several commodity devices being released within the last few years. The Oculus Rift, HTC Vive, and Windows Mixed Reality HMDs like the Samsung Odyssey offer higher resolution screens, more accurate motion tracking, and lower prices than earlier HMDs. They also include motion-tracked handheld controllers meant for navigation and interaction in video games. Because of their popularity and low cost, medical volume viewing software that is compatible with these headsets would be accessible to a wide audience. However, the introduction of VR to medical volume rendering presents difficulties in implementing consistent user interactions and ensuring performance.

Though all three headsets require unique driver software, they are compatible with OpenVR, a middleware that standardizes communication between the HMD, the HMD’s controllers, and VR software. However, the controllers included with the HMDs each has a slightly different control layout. Furthermore, buttons, triggers, touchpads, and joysticks that share the same hand position between devices do not report values to OpenVR in the same way. Implementing volume rendering functions like clipping and tissue density windowing on VR controllers could improve the user’s experience over mouse-and-keyboard schemes through the use of tracked hand and finger movements. To create a control scheme that is compatible with multiple HMD’s A way of mapping controls differently depending on the device was developed.

Additionally, volume rendering is a computationally intensive process, and even more so when rendering for an HMD. By using techniques like GPU raytracing with modern GPUs, real-time framerates are achievable on desktop computers with traditional displays. However, the importance of achieving high framerates is even greater when viewing with a VR HMD due to its higher level of immersion. Because the 3D scene occupies most of the user’s field of view, low or choppy framerates contribute to feelings of motion sickness. This was mitigated through a decrease in volume rendering quality in situations where the framerate drops below acceptable levels.

The volume rendering and VR interaction methods described in this thesis were demonstrated in an application developed for immersive viewing of medical volumes. This application places the user and a medical volume in a 3D VR environment, allowing the user to manually place clipping planes, adjust the tissue density window, and move the volume to achieve different viewing angles with handheld motion tracked controllers. The result shows that GPU raytraced medical volumes can be viewed and interacted with in VR using commodity hardware, and that a control scheme can be mapped to allow the same functions on different HMD controllers despite differences in layout.

Copyright Owner

Jordan King Williams

Language

en

File Format

application/pdf

File Size

79 pages

Share

COinS