Open hinerm opened 8 years ago
Hi there, many of the algorithms are pretty generally applicable, such as Pearsons correlation, Spearman rank correllation etc. 2D histograms are also generally applicable Some of the stuff is specific to 2 channel fluorescence microscopy colocalisation, such as Costes auto threshold and Manders' coefficients, which build on Pearsons correlation. Maybe we need to decide what to move to ops and what to keep in a specific coloc_3 plugin. it was always the long term plan to move generally usefull stuff from Coloc_2 into a more generally useful place, and i guess ops is now that place. So Im all for the above idea.
@etarena good news indeed! Pearsons r, and the two kinds if rank corellearion coefficient are probably the most general nuts if maths suitable for port to ops. Manders' coefficients are typically domain specific to cell biology but could be applied to other images.
The tricky bit is probably the thresholding of images which requires quite sophisticated implementation to be able to do the studd in the Costes paper. We implemented in coloc2 max a d min thresholds (so 2 threshold per channel) and a way to decide if they are exclusive or inclusive... Or at least tried to do what was correct in that respect.
I suppose there should be or maybe already is some image thresholding implementation in ops...and/or imglib2.l, as well as the old IJ threshold tool??? I will leave it to the experts to figure out where to put or merge this relatively fancy thresholding stiff from coloc2 into
Here I just.make a note of what I remember we did in coloc2 and why.
We want to expose colocalisation functionality in ImageJ Ops. This way individual algorithms can be reused in other contexts (e.g. KNIME), and traditional ImageJ/Fiji functionality will benefit from Ops extensibility.