This collaborative research brings together computer scientists from University of Southern Mississippi (USM) and Brown University and neuroscientists from the University of Mississippi Medical Center (UMMC) to study the design of a scientific visualization language (SVL). Despite the numerous visualization approaches already devised, visualization remains more of an art than a science. Grounded in theories and methods from human-centered computing, machine learning, and cognitive psychology, this work is to develop and evaluate a scientific visualization language (SVL) to provide a principled way to help scientists understand how and why visualizations work. Tools and theories developed in this project can lead to efficient knowledge discovery to help neuroscientists study brains using diffusion tensor magnetic resonance imaging (DTI).
This work has the following specific objectives and outcomes: (1) close collaboration with scientists to discover, refine, and verify a symbol space, (2) a semantic space that describes the relationship among symbols, (3) a testbed that implements SVL for neuroscientists to compose visualizations, (4) development of new and enhanced courses at University of Southern Mississippi and Brown University, and (5) wide dissemination of the research outcomes through open-source software, experimental data, open labs, publications, and presentations.
This project is expected to have broad impact. It may lead to significantly better approaches to human knowledge discovery and decision making in many disciplines where visualizations have found successful application, including neuroscience, biomedicine, bioinformatics, biology, chemistry, geosciences, business, economics, and education. Undergraduate and graduate students are expected to participate in the research through our courses, and student exchanges are planned between USM and Brown. K-12 students can visit the USM lab while the project is in progress. Software and results will be disseminated via the project Web site (https://sites.google.com/site/simplevisualizationlanguage).
In this multi-site collaborative project, researchers at Brown University, at the University of Southern Mississippi, and subsequently at the University of Maryland Baltimore County worked to advance the state of the art in creating visualization software tools for scientific users. In particular, tools for studying complicated medical imaging data of brains were a driving applications. This report attempts to focus on the work done at Brown. Much of the impact of this work is in a better understanding of effective practices for creating visual representations of scientific data. This includes experimentally gathered feedback on the value of different ways to construct, display, light, and shade complex 3D models of the brain, including the value of interactive 3D stereoscopic visual analysis. Also included are experimental evaluations of abstract representations of brain connectivity, of scatterplots, and of other scientific information display. A second area of impact of this work is in a better understanding of how to evaluate scientific tools. Generally speaking, it is difficult, time consuming, and expensive to experimentally evaluate tools via expert user feedback. To reduce that expense and time, we have developed and used some evaluation methods that can, in some cases, use less expensive crowd sourcing and modeling to achieve similar evaluation results. Ultimately, these advances in understanding both of specific visualization approaches and of methods for efficiently testing and evaluating visualization approaches will lead to more effective visualization tools for scientists to use in their study of brains and other complex topics. The ultimate impact will be an acceleration of the pace of science, with all the benefits that that can bring.
