BrainSCANr’s circular graph has gotten a face lift and some new features. Jess has been busy the last few weeks adding some interactive elements to the circular graph. We’ve tossed the force directed graph in favor of a more geometric hyper tree. Visitors can still move nodes around by dragging and dropping them into new positions, but we’ve also added some new interactions as well.
Filter by category
Click a category in the circular graph’s key to remove that category from the graph.
Double-click to search
Double click the label of a node to search brainSCANr for that node. Note: you have to double click the label and not the node or it won’t work (we’re working on it).
Hover for clarity
Hover your mouse over a node to see that node’s connections highlighted as well as that node’s probable association with the search term and category. You will also see that node’s corresponding bar highlighted in the new...
...Histogram of probable association!
To better visualize the relative probable association of terms to the search term we’ve added a bar chart histogram to the circular graph. Hover your mouse over a bar to see the term, and probable association, you’ll also see the that term’s node highlighted in the circular graph.
Nice additions.
ReplyDeleteI've noticed two slightly annoying quirks of visualization. First, when I hover over a node to see the connections, I need to be careful not to also hover over a label. If the node/label is on the left side of the graph, then the box containing the category & words covers most of the graph. Perhaps there's a way to make that box appear in a location to minimally cover the rest of the graph?
Also, if I pull a node far away from the circle to better highlight it's connections, the edges curve to a common inflection point before connecting to other nodes. I don't know if this is a standard feature of this type of graph, but it defeats the purpose of pulling nodes away to clarify structure.
Dan: The legend can be dragged and moved away. Anywhere on the screen, actually.
ReplyDeleteAs for the second point... that's... a good one that I don't have a solution to yet. We'll check on it.
Actually, I think Dan is talking about the roll over box that includes the number value for the probable associations for each node, not the legend. The roll overs do obscure the graph if you're rolling over the nodes on the left.
ReplyDeleteThe common inflection problem is also a valid criticism. Both of these are default settings for this particular graph in the JavaScript library we're using. I'll definitely look into modifying them, thanks for the comment!
learned about this from your poster at SFN and i'd like to let you know that you did an a great job explaining it and I'll be looking into using it when starting future work. my question is about how this could relate to the connectome.
ReplyDeleteAre you in the Sebastian Seung camp and believe , as he would say, that you are your connectome? or do you see your tool to get us to the connectome and then we'll all move on the next great frontier in neuroscience whatever that may be?
Thanks, Kelson! I don't think that we "are our connectome" any more than we "are our DNA". These are *very important elements*, to be sure, but it's more complicated than that.
ReplyDeleteThe "connectome" aspect of brainSCANr is mostly just a useful tool... a way to get a quick snapshot of what's happening. The crux of what it can do is the semi-automated hypothesis generation.
Anyway, I've written a bit on the connectome/brain simulation stuff, if you're interested in reading more.
http://blog.ketyov.com/2010/11/cargo-cults-of-brain.html
http://blog.ketyov.com/2011/11/rest-is-just-details.html