Title: Visualizing Relationships in Uncertain MR Spectroscopy Data

Speaker: David Feng, University of North Carolina

Date/Time: May 10, 2010, 11:00 am       

Location: CSRI Building/Room 90 (Sandia NM)

Brief Abstract: Conveying data uncertainty in visualizations is crucial for preventing viewers from drawing conclusions based on untrustworthy data points. I will be discussing a method for augmenting standard multivariate plots like the scatter plot and parallel coordinates plot to incorporate statistically modeled uncertainty.   These new density-based plots preattentively draw viewers to identify values of high certainty while not calling attention to uncertain values. Computing high quality density plots can be expensive for large data sets, so I will also describe a probabilistic plotting technique that summarizes the data without requiring explicit density plot computation.

The driving application for these techniques is magnetic resonance (MR) spectroscopy, by which Radiologists capture metabolic spectra on a 3D grid in the body.  A statistical fitting routine estimates the contributions of individual metabolites to these spectra, resulting in per-voxel metabolite concentrations that are normally distributed. The goal of the visualization is to help radiologists identify relationships among these different metabolites that distinguish between healthy and diseased tissue.  I will show how this technique has succeeded in helping radiologists to identify such relationships for brain tumors and prevents them from drawing incorrect conclusions based on uncertain values.

CSRI POC: Andy Wilson, 505-844-1089