Evaluating Visualisations and Automatic Warning Cues for Visual Search in Vascular Images

Background

In the following sections, we examine visual search in the area of medical imaging, more specifically, 3D vascular imaging. Vascular disease diagnosis can be done effectively by means of 3D imaging techniques such as Magnetic Resonance Angiography (MRA) and Computed Tomography Angiography (CTA). The most common vascular diseases that these imaging techniques help diagnose are stenoses (abnormal narrowings in blood vessels) and aneurysms (abnormal widening or ballooning of blood vessels). We focus on Contrast Enhanced MRA as this is the area of our current research (Suinesiaputra, et al., 2009). CE-MRA involves injection of a contrast agent in the blood stream, making the relevant blood vessels show up as high-density areas on the MRA scan. The output of CE-MRA is a 3D density field (typically around 128x128x128 pixels), with higher densities representing blood flow.

With help of 3D imaging techniques, the thickness of the inside of the vessel (the vessel lumen) can be determined precisely, and assessed quantitatively. This is a very attractive tool in diagnosis. Determination of the vessel lumen is called segmentation. It results in a 3D (usually tubular) surface representing the vessel lumen circumference. Since manual segmentation is time-consuming, automatic segmentation is proposed (Boskamp, et al., 2004; Suinesiaputra, et al., 2009). Automatic segmentation typically first constructs a pathline, which is a line that goes through the lumen of the vessel to be segmented (Boskamp, et al., 2004). From the pathline, a triangle mesh is constructed indicating the lumen. The pathline is also useful as a basis for “intelligent” visualisations.

In automatic segmentation, occasional errors are currently inevitable. Therefore, clinicians have to detect (and correct) errors in the automatic segmentations by means of visual search. This visual search task is the task that we examine in this chapter.