I lead the Visual Analytics and Communications Lab (VACLab for short) where we study and develop visual methods for information analysis and communication. This page includes a gallery of our work, as well as descriptions of projects I completed prior to my arrival at UNC in January 2014. You can click on a project name to learn more about a specific piece of work.
Meanwhile, we’re also working on handful of exciting new projects that haven’t matured enough to post here. Watch this space to learn more as our work progresses! Better yet, join us and become part of the team!
We’ve developed visualization-based interfaces which couple closely with machine learning algorithms to aid in this type of task. This includes work with deep learning models for heart arrhythmia detection which won an Honorable Mention Award at ACM CHI 2018.
We are designing, developing, and evaluating a set of Contextual Visualization Methods for exploratory data analysis which are designed to support the discovery of more robust and generalizable insights from high-dimensional data.
In this project, we explore novel visualization methods for clustering event sequences into threads based on tensor analysis. The threads are then visualized to display latent stage categories and evolution patterns. Interactive grouping allows threads to be grouped by similarity into time-specific clusters.
The American Medical Informatics Association (AMIA) Visual Analytics Working Group (VIS WG) Task Force on Evaluation (TFoE), chaired by Dr. Gotz, is developing a report to (1) describe the state-of-the-art in visual analytics evaluation within the healthcare domain, and (2) provide a recommendations for best practices in future medical informatics research activities.
In this project, we develop a Sankey diagram-based visualization design to visually summarize care plans based on our medical collaborators’ requirements. Our tool is applied to a medical dataset of pneumonia patients collected from a Children's hospital in Shanghai, China.
This project explores how visual analytics methods can be used to create interfaces which interactively measure and visually convey to users the degree of drift in representation during iterative visual cohort selection.
Data with multiple uncertain labels are common in many situations. However, existing visualization approaches often either discard the uncertainty information, or map the data to a low-dimensional subspace where their associations with original labels are obscured. We propose a novel visual technique, UnTangle, for visualizing such uncertain multi-labels.
As datasets grow and analytic algorithms become more complex, the typical workflow of analysts launching an analytic, waiting for it to complete, inspecting the results, and then re-launching the computation with adjusted parameters is not realistic for many real-world tasks. In this work we explore an alternative workflow, progressive visual analytics, which enables an analyst to inspect partial results of an algorithm as they become available and interact with the algorithm to prioritize subspaces of interest.
We explore interactive visualization and data mining techniques to uncover frequent event patterns that have strong associations with user-defined outcome measures. Though our methods can apply generally to a range of temporal event analysis problems, we focus primarily on electronic health record data and related health informatics applications.
A visual analysis method for exploring the differences between (a) official treatment guidelines and (b) the treatments actually performed on patients. A prototype implementation has been developed to understand guideline deviations for a population of cancer patients at a major European hospital.
We propose a method for integrating visual data exploration and on-demand data analytics with an intuitive drag-and-drop user interface which allows clinicians to iteratively refine a group of patients for cohort selection.
Outflow provides a unique flow-based visaulization environment for exploring aggregate event sequence data. The method provides intuitive visual encodings for event sequences, timing, and associations with outcomes. Our approach also helps uncover confounding factors within high dimensional event spaces. We've applied Outflow to applications in healthcare and beyond.
The Intelligent Care Delivery Analytics (ICDA) platform is a prototype system which enables a broad range of risk assessment analytics that process large collections of dynamic electronic medical data to identify at-risk patients.
DICON is an icon-based cluster visualization that embeds statistical information into a multi-attribute display to facilitate cluster interpretation, evaluation, manipulation, and comparison.
SolarMap is a multifaceted visual analytic technique for visually exploring topics in multi-relational data. SolarMap combines several visual techniques including topic contour clusters and interactive multifaceted keyword topic rings.
FacetAtlas is a multifaceted visualization technique for visually analyzing rich text corpora. FacetAtlas combines search technology with advanced visual analytical tools to convey both global and local patterns simultaneously.
This project introduces a new technique called Dynamic Voronoi Treemaps (DVT). Our method's layout algorithm is specifically designed to support smooth, real-time animation of time varying hierarchical data while maintaining desirable aspect ratios, overcoming many key limitations of prior Voronoi-based Treemap methods.
Behavior-Driven Visualization Recommendation (BDVR) is a novel approach to visualization recommendation that monitors user behavior for implicit signals of user intent to provide more effective recommendation.
HARVEST is an intelligent visualization system designed to empower everyday business users to derive insight from large amounts of data. It combines three key technologies to support a complex, exploratory visualization-based analysis process without requiring users to be visualization or computer experts.
The InsightFinder and ScratchPad tools are extensions to the standard browser interface that are designed to capture, organize, and exploit the information discovered while performing a sensemaking task. In addition to providing a graphical note taking space with enhanced bookmarking capabilities, the technologies include context-aware algorithms that help connect portions of a user's notes to relevant portions of a webpage.
Insight provenance—a historical record of the process and rationale by which an insight is derived—is an essential requirement in many visual analytics applications. To develop a model of insight provenance, we characterize users’ visual analytic activity at multiple levels of granularity. This characterization is driven by user studies of analysis activity in various domains, ranging from travel planning to intelligence analysis.
The Channel Set Adaptation (CSA) project provides a framework that allows for the efficient streaming of non-linear datasets to large user groups.
IRW is a new representation for image-based interactive walkthroughs which exploits spatial coherence and rearranges the input samples as epipolar images.
PixelFlex is a multi-projector display system that combines multiple roughly aligned projectors into a unified high-resolution display. It includes an automated camera-based calibration process and two rendering algorithms that support interactive applications.
The NAVE (Non-Expensive Automatic Virtual Environment) is a stereoscopic, multi-projector display system built from off-the-shelf components.
This project—a joint study between Georgia Tech, Emory University, and the VA Hospital in Atlanta—developed a virtual reality-based exposure therapy system to treat Vietnam veterans suffering from post-traumatic stress disorder (PTSD).