[brainCOLOR] SfN Poster on protocol color?
Andrew J. Worth
I would like to propose that we present a poster at the Society for
Neuroscience Meeting in San Diego (Nov. 13-17) on the topic of colors
for brain labeling. I would like to be able to give away printed
posters that show the current state of the cortical parcellation
protocol we are developing. It would also be cool to have a live demo
on a laptop where people could play around with the colors and see the
results on a fully labeled brain in 2 and 3D.
Please let me know if you are interested in being on the poster or even
taking the lead on the poster or if you have any comments or
suggestions. The abstract must be sent in by the 13th (one week from
today).
Andy.
Bennett Landman
these compare with state-of-the-art automated solutions - FreeSurfer+Cruise?
It would be great to show how much room there is for improvement! Too many
people - especially in neuroscience - consider automated parcellation a
"solved" problem.
Cheers,
Bennett
Andrew J. Worth
the demo, we can use one of the parcellated scans that looks "typical"
at the time. We're still developing the parcellation protocol and the
labeling software so we're not quite ready to do that manual vs.
automated comparison. We'll only be ready for that after we've done a
few hundred scans.
It is easy to be seduced by freely available automated methods that work
pretty well.
Andy.
Bennett Landman
run the programs on some data and you can see if 1 example of each is worth
showing? I'm just thinking that it might make a bit more impact than if it's
presented without "motivation."
Arno Klein
bennett --
by automated solutions, you are referring to automated labeling, not automated coloring of labels. automated parcellation, which i agree is far from a solved problem, isn't relevant to the coloring problem -- it doesn't require colors at all except when visualizing the results, whereas coloring is critical when manually labeling brain image data. what makes the coloring problem interesting to me is that it is practical, visual, and quite general but happens to be particularly challenging in a convoluted structure like the brain.
andy --
i would be very excited to be involved in this project and would be happy to take the lead except that i'm involved in three grant applications that are due internally this week! would you be able to write up a draft incorporating some of the ideas that have been brought up by the group?
cheers,
@rno
Do you have the brain that you would like to show? Perhaps we can show how
these compare with state-of-the-art automated solutions - FreeSurfer+Cruise?
It would be great to show how much room there is for improvement! Too many
people - especially in neuroscience - consider automated parcellation a
"solved" problem.
Cheers,
Bennett
-----Original Message-----
From: brain...@googlegroups.com [mailto:brain...@googlegroups.com] On
Behalf Of Andrew J. Worth
Sent: Thursday, May 06, 2010 11:30 AM
To: brain...@googlegroups.com
Subject: [brainCOLOR] SfN Poster on protocol color?
Arno Klein
i do like the idea of an interactive color mapping demo.
perhaps individuals could toggle factors in a cost function that affect the color mapping, something along the lines of:
1. psychophysics criteria for discernibility (luminosity, hue, contrast, etc.)
2. color chart dynamic range and distribution
3. neighborhood relations (clustering of colors into spatial groups, etc.)
what would be particularly cool, and actually useful, but most probably out of scope, would be to define particular tasks that an individual would attempt using different color mappings (short of labeling brain region boundaries). there are a few examples i am aware of in the information visualization literature where different representations of data improve or detract from the ability to perform a task on those data. for example, try estimating node density in or shortest path through a network given (a) an associative matrix vs. (b) a node-link diagram: each has its advantages and disadvantages depending on the context. in our case, we're changing the color map, not the form of the data, so we could probe far more subtle effects.
cheers,
@rno
Bennett Landman
We talked offline. I now see the logic in waiting for a full comparison between the manual protocol and the automated methods.
Cheers,
Bennett
From: brain...@googlegroups.com [mailto:brain...@googlegroups.com] On Behalf Of Arno Klein
Sent: Sunday, May 09, 2010 10:17 PM
To: brain...@googlegroups.com
Bennett Landman
-Bennett
Arno Klein
This project interests me too much to let such an opportunity slip by --
Andy has sent me a draft of the abstract. I will work on it further right now. If anyone else wants to contribute, give a holler, because it's due at 5pm!
cheers,
@rno
On May 12, 2010 5:08 PM, "Bennett Landman" <bennett...@vanderbilt.edu> wrote:
This would be really cool. However, I don't have the bandwidth for sfn this year... If anyone wants to lead, I'll support.
-BennettSent from my iPhone
On May 9, 2010, at 10:28 PM, Arno Klein <binary...@gmail.com> wrote:
>
> i do like the idea of...
Bennett Landman
-Bennett
Arno Klein
--------
“An interactive tool for choosing optimal brain colormaps”
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,
number of colors, etc. 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 color schemes on fully labeled MRI brain scans where the user can weight factors in a cost function that affect the color mapping. 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.
--------
cheers,
@rno