David Thomas Gering

Graphlets: a method for visualizing dynamic data

Dynamic contrast-enhanced MRI is often performed to characterize tissue by observing the passage of an agent injected into the bloodstream. MR images are acquired at regular intervals during a time period that begins prior to injection of the contrast agent, and extends through the agents passage through the vasculature under study. The MRI signal observed during this time period can be transformed into a curve of contrast concentration over time [Ostergaard]. Parametric maps can be computed from the time-series and overlaid on an anatomical image for review by radiologists. Typical hemodynamic parameters include regional cerebral blood flow (rCBF), mean transit time (MTT), and regional Cerebral Blood Volume (rCBV). The shortcoming of this approach is that the parametric maps, although convenient, do not allow the clinician the opportunity to observe the full richness of the dynamic data.

Flattened anatomy for interactive segmentation & measurement

Anatomical surfaces can be extracted from a volume of medical imagery through segmentation, which is the process of labeling image voxels according to the tissue type represented. Many anatomical surfaces can be described as 2-D manifolds embedded within 3-D space. The manifolds could be linear, such as a plane, or non-linear, such as a curved sheet. In some applications, it would be desirable for the user to be able to interact with the manifold by drawing upon it in some way. The purpose of this drawing could be to perform quantitative measurements such as to measure distances, surface areas, or volumes. The purpose could also be to analyze local properties of the 3-D surface at specific locations, where these properties include thickness and curvature.

A flight simulator approach to the visualization of dynamic medical data

Visualization of 3-D medical scan data (e.g.: MRI, CT, PET) is often performed using surface rendering or volume rendering. Volume rendering is well suited for applications where the assignment of color and opacity values is straightforward given the image voxel intensities. This is true of CT data where there is a strong correlation between Hounsfield units and tissue types. Surface rendering is well suited for applications where segmentation of key structures is a necessary step for good visualization, such as with MRI data, and a polygonal mesh can be wrapped around segmented structures to form a surface model.

Integrated image zoom and montage

Standard features of medical image viewers often include zoom & pan, image montage, and blended overlays. Zooming and panning are typically implemented with a convenient real-time mouse interface. For example, dragging while pressing the left mouse button pans, and dragging while pressing the right mouse button zooms. A montage refers to constructing a tiling, or 2-D array of images, from a set of images acquired at various slice locations or time-points. An overlay refers to blending one image with another as a form of image fusion. A parameter, alpha, specifies the opacity of a foreground image that is overlaid on a background image.

Interactive animations for medical data visualization

For many applications of 3-D medical data visualization, common graphics hardware is capable of performing surface rendering of a few segmented structures in true real-time. Volume rendering, which can generate photo-realistic images, can be performed in near real-time by making a trade-off in image quality. An example of this trade-off is the feature where medical images are rendered with reduced resolution while a user is rotating the viewpoint using the computer mouse or other input device. Then, when the user pauses his/her movement, the computer spends more time, such as a couple seconds, to render the scene at a higher resolution.