HealthHack 2015 Challenge
Neuroimaging datasets are large, too large for desktop computers, the answer is the web. We can currently view multi GB and TB imaging datasets via a HTML5 + JavaScript interface based around tiling in a Canvas Element. (click the image for a link to the real interface)
A detailed overview of the interface and how it works can be seen on YouTube
The sizes of the various types of data can be seen below:
Note that all of this data is 3D and thus the interface will only show a current tri-planar view through the planes of the volume
Our (international) users have now become very familiar with this style of interface but want more. The #1 request we have is for annotation. A mocked up example of this is shown below:
Even the ability to add simple elements on top of the image would be a great outcome. Landmarks, simple polygonal lines and text.
How this data is stored is a vexed question, currently all data on a TissueStack server is open, so the tracings and annotations would also be open. For now everything being open is a reasonable approach.
Thought needs to be given as to how tracings and annotations work beyond a 2D plane as is more typical in traditional web interfaces. A common approach it to treat a point as a sphere and show smaller circles in adjacent slices. Failing that a user will only be able to find annotations by chance when scrolling through the volume. User cues should feature heavily here but the mechanism for doing this requires a group of coders with some novel visual design ideas for web interfaces.
Collaborative annotation across multiple instances of TissueStack. We have diverse groups of researchers working from multiple sites in Australia (and internationally). Ideally each volume should be traced by the combined knowledge and skill of multiple experts. Being able to collaboratively annotate images would be a fantastic boon to science ala google docs and other collaborative tools.