ISRG Schedule, Fall 2009

Discovering and extracting linear trends and correlations in datasets is very important for users to understand multivariate phenomena. However, current widely used multivariate visualization techniques, such as parallel coordinates and scatterplot matrices, fail to reveal and illustrate linear relationships intuitively, especially when more than 3 variables are involved or multiple trends coexist in the dataset.
We present a novel multivariate model parameter space visualization system that helps users discover single and multiple linear patterns and extract subsets of data that fit a good model. Using this system, analysts are able to explore and navigate in model parameter space, interactively select and tune patterns, and refine the model for accuracy using computational techniques. We build connections between model space and data space visually, allowing users to employ their domain knowledge during exploration to better understand the patterns they discover.

This talk is a preparation for my proposal which will take place in the very near future. The topic is Previsualization in which I use a combination of Machine Vision and Computer Graphics to sense, interpret and visualize a scene. Here is part of the abstract from my proposal:

Previsualization is a class of techniques for creating previews of a movie sequence in order to visualize a scene prior to shooting it on the set. Previous methods for Previsualization have employed a wide variety of different media to compose the artistic and logistic elements of a scene, including hand-drawn storyboards, scale models, and photographs. Despite being quite prevalent in movie-making, these methods are inherently limited. For example, each of these methods can take long periods of time to create, may require teams of artists to complete, and lack interactivity. A recent trend in Previsualization is to use video game engines to add interactivity to previous methods. However, game engines also have similar limitations because they require time consuming and complicated manual reconstruction and importing of sets. Previsualization can be used to incorporate many aspects of the scene, such as sets, dialogue, action sequences, and special effects. For this project the focus will be on a new method for Previsualization of the objects within a scene in order to facilitate re-lighting.

Patient motion during lengthy medical imaging procedures can severely degrade the accuracy of diagnoses made from the resulting inaccurate reconstructions of the patient’s organs. Motion correction algorithms have been and continue to be developed in order to solve this problem. Frequently, testing of these algorithms is done either by using simplistic deformations that do not accurately simulate true patient motion, or by applying the algorithms directly to data acquired from a real patient, where the true internal motion is unknown. This presentation describes a way to combine these two approaches by building numerical models directly from MRI's or volunteers in different configurations, thus capturing realistic the real motions that happen in the body. These phantom can then be used to simulate the data that the motion correction algorithms would normally observe, allowing direct comparison between the motion corrected reconstructions and the true organ configurations used in the projections.

Last semester I presented our proposed visualization techniques to convey the change of data patterns in multivariate data streams. They include superimposition, juxtaposition, step juxtaposition, and animation. In order to convince visualization researchers that our techniques are useful and effective, we must show that they can make the extraction of data pattern change easier compared to traditional time-series data visualization techniques, such as line charts and heatmaps. In addition, we want to develop guidelines to advise analysts how to choose visualization techniques based on the characteristics of datasets and tasks. The above issues are two goals of the user study discussed in this talk. In this ongoing user experiment, we chose two common data patterns, linear trends and cluster positions, and then created some streaming datasets, each of which has changes with various magnitudes on one of these two types of patterns. Participants are asked to determine if and how data patterns change over time. Based on some preliminary experiment results, we will analyze the capabilities for each visualization technique for conveying the change of data patterns, as well as derive a guide to help analysts choose techniques based on their datasets and tasks.

TBD

TBD