I’m new to CellProfiler and loving it. Yet, as a newb I’m running into issues. I’ll try to be clear with my question. Please ask me to be more specific if not clear. To avoid revealing too much info, I would like to not share the pipeline or image set unless I can do so in private.
My question is: when tracking cells and cell components (nuclei, cell borders, cytoplasms, etc) as primary, secondary, and tertiary objects) over 150 frames or so, or even a few, is there a way to prevent or outlaw merging and splitting of the cells or cell components, as well as potentially outlawing new objects (although this can be tolerated and managed post-processing), without using the LAP tracking method?
My problem is: I’ve tried the overlap and distance approaches (as well as some other things), but the objects being tracked are highly motile, very homogeneous with respect to fluorescent intensity, and subject to a variety of shape and velocity changes which don’t follow any pattern with respect to neighbors, or even themselves frame-to-frame. Accordingly, when tracking multiple objects, sometimes the distance to the nearest neighbor (also being tracked) is less than the inter-frame (frame-to-frame) movement distance of the object(s) of interest. So, such objects are deemed by the program as splits, merges, or new objects. Increasing the overlap or distance measurements seems to result in a miscall, again a split or a merge with a closer object rather than the one being tracked. Decreasing the distances seems to result in lots of new objects being identified. One or two is ok, but multiple new objects per multiple frames is not able to be handled even post-processing.
My analysis would really benefit by being able to prevent such merges and splits, as the objects at the beginning of the analysis could be retained throughout the session with major hiccups. I’m aware the LAP method can possibly reduce such events drastically or completely. However, I’m avoiding the LAP method because 1) it doesn’t seem to easily enable testing to ensure pipeline parameters, values, etc hold up for a given image set, or tens of image sets, and 2) I need (or find very useful) the ability to obtain quality control images from every frame (with tracking numbers attached on displayed images). This enables me to ensure that 3 or so objects that I’m tracking retain the same tracking label throughout 150 or so time-course images. Essentially- ensuring number 1 object is always associated with number 1 objects throughout an image set, and so on for 2, 3, etc…up to about 10 objects over 150 - 200 frames. Obtaining these frame-to-frame tracking numbers displayed on images for each frame doesn’t seem possible using LAP as all labels are re-assigned at the end of the analysis via post-processing. If such a modification could be made to negotiate this problem using only CellProfiler it would be great. It might also be helpful to not have to use LAP and CellProfiler Analyst to prevent splitting and merging.
All that being said, is there a way or is the only option to use LAP and CellProfiler Analyst?
Pardon the verbosity.