GRACE: A Visual Comparison Framework for Integrated Spatial and Non-Spatial Geriatric Data

• Published in: IEEE transactions on visualization and computer graphics 2013-12, Vol.19 (12), p.2916-2925
• Main Authors: Maries, Adrian, Mays, Nathan, Hunt, Megan Olson, Wong, Kim F., Layton, William, Boudreau, Robert, Rosano, Caterina, Marai, G. Elisabeta
• Format: Article
• Language: English
• Subjects: Aged; Aged, 80 and over; Algorithm design and analysis; Algorithms; applications of visualization; Biomedical imaging; Brain modeling; Computer Graphics; Data visualization; Decision Support Systems, Clinical; Design engineering; Design studies; Diagnosis, Computer-Assisted - methods; Female; Geriatric Assessment - methods; Geriatrics; GRACE (experiment); high-dimensional data; Humans; Imaging, Three-Dimensional - methods; integrating spatial and non-spatial datavisualization; Interactive; Male; Mathematical analysis; Mathematical models; methodology design; Rendering (computer graphics); Reproducibility of Results; Sensitivity and Specificity; Software; Studies; Systems Integration; task and requirements analysis; User-Computer Interface; Visual; visual comparison; Visualization
• ISSN: 1077-2626; 1941-0506
• DOI: 10.1109/TVCG.2013.161

We present the design of a novel framework for the visual integration, comparison, and exploration of correlations in spatial and non-spatial geriatric research data. These data are in general high-dimensional and span both the spatial, volumetric domain - through magnetic resonance imaging volumes - and the non-spatial domain, through variables such as age, gender, or walking speed. The visual analysis framework blends medical imaging, mathematical analysis and interactive visualization techniques, and includes the adaptation of Sparse Partial Least Squares and iterated Tikhonov Regularization algorithms to quantify potential neurologymobility connections. A linked-view design geared specifically at interactive visual comparison integrates spatial and abstract visual representations to enable the users to effectively generate and refine hypotheses in a large, multidimensional, and fragmented space. In addition to the domain analysis and design description, we demonstrate the usefulness of this approach on two case studies. Last, we report the lessons learned through the iterative design and evaluation of our approach, in particular those relevant to the design of comparative visualization of spatial and non-spatial data.