University of Oregon directed research project ~ Spring 2015
|
This project is one of the major milestones in the process of completing a PhD at the University of Oregon. It is designed to show that a student knows how to, and can, conduct and then present on research and results. I am happy to report that after completing my presentation my committee recommended pass, and I have now moved from a Conditional Doctoral Student, to a Doctoral Student!
My presentation slides are attached, and a few interesting images from our new visualization technique for level-of-detail tractography visualization are included here as well. |
| ||
Graduate Research Forum 2014
|
First presentation on phd researchThis presentation went very well. There were over 100 poster presenters from all over campus, and many from the University and local community came out to support the graduate research being done here at the University of Oregon.
The work that I presented was my work with managing extreme scale tractography data for effective visualization: Introduction Tractography is a 3D visual representation of neural paths used in research and clinical environment. Advancements in sensing methods have allowed data to be collected at resolutions higher than ever before, often producing up to 750 GB of data per person. Existing equipment and methods are ill-equipped to handle the scale of this problem, often utilizing under 10% of the data, and when fully displayed the visualization is visually unmanageable. This leads to the need to holistically understand and reduce the data so that more advanced visualization methods can be used. Data Reduction The base form of this tractography data is a polyline consisting of an average of 500 individual points per tract. We implemented two subsampling methods for polylines, nth-element and colinear points, in order to reduce the number of individual segments in individual tracts. The first method is a naïve approach which only keeps every nth point in a polyline. This method gave great data reduction percentages for the subsampled tracts over the original tracts, but gave an unacceptable level of error since important features of the data were being ignored in the subsampling. To address the large error of nth-element subsampling, colinear points was used. Colinear points is an approach that will get rid of a point from a polyline if it lies within a given tolerance on the same plane as the previous and next points. Using this approach, important data features such as curves are able to be kept, while straight segments can be aggressively pruned. This approach was able to achieve a good level of data reduction with very minimal error. Summary Data reduction is a critical aspect in enabling effective exploration of extreme scale tractography data. With accurate and effective reduction techniques minimal error can be introduced into the data set while the overall size becomes manageable with advanced visualization techniques that allow specific regions of interest to be show. | ||
