Abstract: In this work, we explore issues of constructing and using an optimal colormap for labeling and visualizing brain anatomy. To this end, we will create an interactive brain colormap tool and will apply this tool in one of our ongoing projects, the definition and use of a labeling protocol to manually label 1,000 MRI brain scans, an intensely visual task (see www.braincolor.org).

We propose a psychophysics solution that considers criteria for discernibility (perceptual differences of luminosity, hue, contrast, etc.). We will construct colormaps based on weighted combinations of constraints, including: maximal perceptual difference between any two labels, minimal chrominance difference between any two labels within a set of labels (considering neighborhood relations), color chart dynamic range and distribution, and number of colors. We are also interested in constructing colormaps that have esthetic value and can provide additional information such as hierarchical grouping of regions while remaining distinct across anatomical boundaries.

We will create an interactive color selection demonstration showing different colormaps on fully labeled MRI brain scans, where the user can weight factors in a cost function that affect the colormap. We will compare our colormaps with past color and texture schemes used in brain atlases and labeling systems, and with the use of random saturated colors that emphasize certain borders and not others. Software for creating brain labels and for displaying them need to be able to dynamically choose colors to match the needs of the user or viewer. Our approach is intended to make this selection process more rigorous and principled.