This thesis develops a 3D manipulation approach for MARS spectral CT datasets for medical diagnosis/imaging. The key outcome is the design of a novel 2D/3D hybrid user interface for interactive exploration. A study is presented that shows the effectiveness of the hybrid user interface compared to a standard 2D interface for the novice (medical student) as well as experienced (radiology resident) users in diagnostic radiology. The study demonstrated that the hybrid interface is an effective approach that requires minimal training to achieve consistently accurate results. In fact, using the hybrid interface, the students’ accuracy scores matched that of trained residents.

Spectral CT (also referred to as colour CT) is an emerging medical imaging modality that acquires data over multiple x-ray colours in order to provide more accurate images and expand medical applications. The MARS project has developed a small specimen spectral CT scanner for pre-clinical research. This thesis work builds on MARS Vision, an in-house visualisation tool developed for analysing MARS datasets.

The first accomplishments in this thesis advance the tools and features in MARS Vision. Some of the main features include stereoscopic 3D rendering, rapid mesh extraction, a rendering engine, and an arbitrary slice view. These features also form a framework to facilitate primary research into 3D manipulation. The thesis presents a hybrid user interface that combines the zSpace stereoscopic display (with 3D stylus input) with a standard 2D display (with mouse and keyboard input). The interface augments the diagnostic radiology workflow by adding a 3D component, as opposed to replacing the existing workflow.

This thesis also presents an evaluation of the hybrid interface by comparing it to a standard 2D interface (based on Inteleviewer) along with a 3D only interface. The study involved 21 medical students and 10 radiology residents performing a scoliosis diagnosis task using the three interfaces. It demonstrated that the hybrid interface was an effective alternative for 3D manipulation in medical diagnosis. This interface is implemented in the commercial version of MARS Vision and is currently being used by several pre-clinical research groups worldwide.