Uncertainty in medical visualization: Towards a taxonomy

•We identify relevant sources of uncertainty in the medical visualization pipeline.•We develop an abstract representation of the types of uncertainty.•We categorize the uncertainty types w.r.t. their mathematical characteristics.•We investigate existing state-of-the-art uncertainty visualization approaches.•We identify open uncertainty visualization challenges.

The medical visualization pipeline ranges from medical imaging processes over several data processing steps to the final rendering output. Each of these steps induces a certain amount of uncertainty based on errors or assumptions. The rendered images typically omit this information and allude to the fact that the shown information is the only possible truth. Medical doctors may base their diagnoses and treatments on these visual representations. However, many decisions made in the visualization pipeline are sensitive to small changes. To allow for a proper assessment of the data by the medical experts, the uncertainty that is inherent to the displayed information needs to be revealed. This is the task of uncertainty visualization. Recently, many approaches have been presented to tackle uncertainty visualization including a few techniques in the context of medical visualization, but they typically address one specific problem. At the moment, we lack a comprehensive understanding of what types of uncertainty exist in medical visualization and what their characteristics in terms of mathematical models are. In this paper, we work towards a taxonomy of uncertainty types in medical visualization. We categorize the types in an abstract form, describe them mathematically in a rigorous way, and discuss the visualization challenges of each type and the effectiveness of the existing techniques. Such a theoretical investigation allows for a better understanding of the visualization problems at hand, enables visualization researchers to relate other medical uncertainty visualization tasks to the taxonomy, and provides the foundation for novel, targeted visualization algorithms.

Graphical abstractFigure optionsDownload full-size imageDownload high-quality image (140 K)Download as PowerPoint slide