best to have at least ~50kDa inside the masked area and larger is better, job will work well only if the contents under the mask are decently well-aligned beforehand (not great at collapsing large sweeping motion or compositional heterogeneity). I think it’s wise to try the same job with 4 or 5 masks representing the span between smallest region of interest and entire particle, if any seem to align particularly well then those particles can be used as input for the next, smaller mask job. recenter both shifts and rots after each iteration. make sure in addition to 6 dilation, the masks have padding (soft mask).
