Forums » Appion »
get better particle picking
Added by Anchi Cheng over 13 years ago
From Yen-Chywan,
Attach please find a screen shot of dog picking. You can see some place program pick the wrong area. I can increase the threshold. But then most of the particles in small defocus will not be picked. Can you suggest me what else to tune to pick the right particle in large defocus and do not miss particle in small defocus.
The failed picks are mostly when they are too close to each other or a particular high contrast mass. Dogpicker had difficulty separates correlation peaks in the former case and damp down the really strong mass. I don't think you can expect it or any automatic picker to be perfect without false positives and without false negatives. Changing theshold as you did are the correct way to optimize it. I don't see more you can do in this homogenerous case.
However, if it is a matter of getting better picks, you can use your Dogpicker results, create a stack, align them and classify the 2D images in the stack, and then pick the classes representative to your particles and create template for template picking. It should give you better results.