Collaborative design research: The visualization of medical concepts

Citation

Zender, M., & Crutcher, K.A. (2007). Collaborative design research: The visualization of medical concepts. Proceedings of the International association of societies of design research (pp. 1-24). Hong Kong: http://www.sd.polyu.edu.hk/iasdr/proceeding/papers/Collaborative%20Design%20Research_%20The%20Visualization%20of%20Medical%20Concepts.pdf.

Abstract

The proliferation of data is threatening to swamp our ability to convert data into knowledge. Visualization promises to facilitate this conversion. Yet visual communication designers have not been deeply involved. One potential impediment to involvement is the lack of collaboration between visual communication designers and knowledge workers in specialized domains.

This paper describes a collaborative research project that integrates medical science and visual communication design. The project involves the development of a visual language to represent medical concepts by deriving propositions from papers, breaking propositions into concept objects, designing a visual object system (consisting of icons, glyphs and combinations) to represent the objects, and displaying the objects as a network of concepts with links to the original papers. Prototypes have proven to be highly condensed and accurate yet readable in seconds. If the visualization approach proves successful, the results would be groundbreaking in science and design.

Summary

The problem identified by the authors is as follows: Can key concepts in fields with controlled vocabularies, such as medicine, be efficiently communicated with images such as glyphs or icons, and, if so, are these images able to effectively illustrate the conceptual web surrounding hundreds or thousands of journal articles and papers within a specific area of investigation? If such a system were interactive then it might lead to insights more quickly and if it remained linked to individual papers then the visual display might be an improved means of exploring literature databases such as PubMed (p.4). The authors described and defined the parameters for making a comprehensive visual language for the expression of scientific concepts and contexts.

In developing a visual language, the approach taken by the authors was to “identify key concepts, connect those concepts to summary statements, break those statements into their essential conceptual objects, illustrate those concepts using icons and glyphs, and present these visual objects in an interactive concept space where they could be immediately perceived and understood in relation to each other (p.6).” Graphic forms that serve both icon and glyph function were combined allowing for glyphs to signify category while icons signify the meaning of an object within a category. Families of object icons were designed to mimic the parent/child structure of the UMLS. To depict actions, the authors developed Proposition Statements (“Object A – does something relative to – Object B”), then used thick lines with small graphic representations of action concepts (bind, modulate, produce) to connect Object A and Object B. Upon interaction with the line (roll-over) the objects within the line are animated to better depict the action. All other icons have a tool-tip-like name that pops-up when the mouse hovers for more than one half second.

Methodology

First, the icons of the Visualization Systems were informally evaluated for their communicative quality and then the icon-based display was compared against a similar text-based display. In order to develop a unified Visualization System for formal evaluation, 4 expert reviewers were recruited to rate the communication effectiveness of the icon for each object. A rating scale was used by the reviewers in evaluating the icons, the scores were then averaged, and then icons were redesigned based on the evaluation.

 Figure 1: ApoE beta-amyloid icon evaluation result report page, p. 19

Based on the evaluation of the Visualization System, an experiment was designed to compare the icon-based display and a texted-based display. For the testing, a total of 27 subjects, 13 with icon display and 14 with the text display were recruited. The subjects were a representative population of domain experts who performed tasks to evaluate the displays. The tasks were designed to measure three effects: speed of recognition of concepts; speed of identification of related concepts; speed of identification of the type of relationship between concepts.The tasks for each group were identical. Some required simple recognition and identification while others require interpretation and association.

 Figure 2: The icon-based display test

Figure 3: The text-based display test

Results/Conclusions

Overall, the icon-based display was both faster and more accurate. For simple identification tasks, the two displays were nearly equal in speed of identification. For identification tasks requiring reasoning/association of similar concepts, the icon-based display was overall 18% faster. For the tasks requiring the identification of similar concepts, the icon-based display was nearly twice as fast as the text-based display. Accuracy of the icon-based display was equal to the text-based display on simple identification tasks but far more accurate on tasks that required the recognition of relationships. On task 3, “Count the number of diseases in the display” the icon-based display was 4.43 times more accurate than the text-based display (p.22). While the results are from comprehensive, they are promising in that design and visualizations can effectively communication complex scientific content.

Key Points

  • “One problem is language. In scientific literature, as in most other areas, findings are reported in writing and the concepts are embodied in words. Yet words are often difficult to define, requiring a context to determine their meaning (p.5)”
  • “Icon families mimic the parent/child structure of the UMLS and by doing so created icons with a ‘proximate context’ that enables one icon’s meaning, the parent, to inform the meaning of other icons, the children (p.10)”
  • “In the propositional statements we analyzed, “neuronal degeneration” is one example of a process object: a neuron (thing, noun) degenerates (dies). This would be a conceptual entity in the UMLS.” (Zender and Crutcher, 2007)
This entry was posted in Displays, Systems and tagged , , , . Bookmark the permalink.