CRII: SCH: Multidimensional Tree Diagram Visualization for Linked Data Exploration

Northeastern University, Boston MA

Investigators

Abstract

The analysis and visual exploration of tree diagrams, visual representations of hierarchical data, is particularly challenging in the imaging sciences (e.g., radiology, biology, astronomy, etc.) as the tree diagrams ideally need to be viewed and explored within the context of the original data. The original data from which the tree was derived may be composed of a combination of imaging data cubes (3D volumes), 2D images, and quantitative (tabular) data. A useful technique for the visual exploration of such related data sets is linked views in which equivalent data is visually highlighted in each individual visualization. The goal of the proposed research project is to develop the novel techniques, methods, and taxonomies needed for tree diagrams to enable effective interactive visual data exploration including in the context of multidimensional linked data. The contributions of this proposal are applicable to many domains including and beyond healthcare. The new tools and taxonomies developed as part of this research, along with tutorials and syllabi, will be made freely available and shared through conferences and workshops. The research presented in this proposal will advance the state of the art in visualization through the creation of new techniques, methodologies, and taxonomies for tree diagrams. The results of the research will enable the effective design of interactive tree diagrams in a multidimensional context based on the required analytic tasks in a novel and systematic approach. The full project from taxonomy abstractions to visualization and editing techniques and methodologies to an implemented system with an evaluation case study and user testing provides a model approach to visualization research. The following research questions will be addressed as part of the proposed work: (1) How does one pick the optimal tree diagram type based on data type and analytic task? and (2) What are the best methods for interacting with tree diagrams including how to select data? To answer these questions a new taxonomy for tree diagrams will be created in order to take into account the concepts of data type, data dimensionality, quantitative data encoding options, and task. In order to enable the ability to develop linked views with tree diagrams, a new task taxonomy for tree diagrams will be created including data selection methods for tree diagrams. The taxonomies and related methodologies for tree diagrams will be implemented into an interactive publicly available tool enabling a user to select or invent the most appropriate tree diagram for their data and set of tasks, and then interact with it in a linked-view data visualization and exploration environment for visualizing and analyzing multidimensional datasets. In order to evaluate the new taxonomies and methodologies as well as the usability of the new tool incorporating these concepts, a real world case study and usability evaluation will be conducted in the healthcare field of brain imaging with the goal of investigating and developing a novel tree diagram representation of brain blood vessels.