CP 2.2 vs 2.1 - object tracking

identifyprimary
threshold

#1

I am using version 2.2 on a mac and most of my labmates have 2.1 (mac and PC). We recently realized my results are different than theirs for the same data sets. We are using a fairly basic pipeline with speckle enhancement, primary object identification with a manual threshold and area measurement. Is this a common problem?
Second question; The quantificaiton of the area (in square pixels) changes depending on the intensity threshold for the primary object identification. Is there a way to track this change? In other words can we track the identified objects and their corresponding area measurements? Since CP doesnt label the objects, numbering of the objects in the end file do not help. Thanks!


#2

Hi,

For the first question, my other colleagues may address this better. In my view, it is quite difficult to assure there’s no difference between 2 versions of CellProfiler. You can try compare the discrepancy in statistics (mean, median, sd etc…) and see if the discrepancy really diverse you scientific conclusion?

For the second, the identification of an object should come with Object_Number AND ImageNumber, these two numbers often ensure object identity. So in case you have 2 objects with same object numbers, please look for their ImageNumber as wells.


#3

I wouldn’t expect the results to be different between 2.1 and 2.2; can you post one of your 2.2 pipelines and one of your labmates’ 2.1 pipelines so that we can compare them and see if we can reproduce the difference?

If you’re trying to cross compare cells across analysis runs at different thresholds, probably the easiest way to do it would be to try to match them up by their Location_Center_X and Location_Center_Y; even if the boundaries of the cells change, the centers should stay in largely the same place.

If you wanted to compare the areas systematically across different thresholds, my advice would be to do IDPrimary with your strictest threshold (call them Cells_HighThresh, for example), then use those objects as seeds for IDSecondary as many times as you want for your lower thresholds (Cells_MedThresh1, Cells_MedThresh2, etc, all seeded from Cells_HighThresh)- then you could use the Parent column to always make sure you’re comparing the same cells across thresholds.

You can also use the Workspace Viewer (CP 2.2 and up) or ConvertObjectsToImage or DisplayDataOnImage to try to backtrack the labels of the objects.