acardona
commented
8 years ago In addition, the review system as currently implemented performs the sequence generation and most filtering on the server side, which is slow and crucially consumes precious resources. Would scale a lot better to query the arbor and review data and, without any processing at all, send it out to the client for all processing, using functions from Arbor.js.
In the last several days, discussions have suggested several filters that could be imposed to generate a list of paths to review in the review widget. These would offer enhanced time savings through targeting of focused proofreading.
1) Review paths between node A and node B, both user-defined. This would presumably have an interface like the path search in the graph widget. This mode would generate a single review path starting at B and going to A (or vice versa). This would be used in cases where it is clear that there is a false merge between two large neurons and one wants to find it.
2) Review only paths with Strahler number between user-defined integers n, m. The principle use we imagine for this would be to focus initial review on the core backbone (e.g. Strahler 3 or higher) to sanity check the main structure of a neuron without going through all the messy small bits.
3) Only review nodes within X microns of the base of their root. The idea here is to use in combination with the Strahler filtering, but for distal tips. For example, I would like to review Strahler 1 and 2 branches, but only those nodes within 3 microns of the 'local root' (where that subarbor merges into its backbone, a Strahler 3 branch).
All of this filtering can easily be done locally, which is extra nice.