2 channel live cell tracking

ilastik

#1

Hi,

I have a sequence of 2 channel tif images taken of live cells. My goal is to use a stable nuclear reporter from the TxRed channel to identify and track nuclei masks and then measure the intensity of a GFP reporter within each nuclei. I run a pixel classifier on the TxRed nuclei and create a hdf5 file of the probabilities. I then run a tracking project that takes the TxRed images as the raw data and the hdf5 file of the probabilities and outputs a CSV_Table using the Plugin export. I then run the same tracking project giving it the same hdf5 file of the probabilities but in this run give it the GFP images as the raw data. I get a second CSV_Table file from this run.

While my results look reasonable, are the tracks and lineages in the TxRed and the GFP CSV files referring to the same cells? I’d like to know if this is a valid approach and if there are optimizations I can create.

thank you,
Mark


#2

Hey @MarkDane, just to clarify, for me:

  1. you have a two channel time series
  2. you are training only on the TxRed channel, in both, PixelClassification and Tracking
  3. you run the Tracking project trained on TxRed with the GFP images as raw data?

The object ids in the time frames should be the same in either case since the same probability map is used. However, I would not expect the tracks to be the same, as the raw data is different in each case and features are calculated using this data.

I guess the right way to do it would be to do 1) and 2) for 3) just for extraction of the features, but discard all tracking information from 3)


#3

You have the right description of my analysis.

My GFP signal in the nuclei is transient so I don’t think it can provide accurate information for tracking. TxRed is stable in the nucleus for all images. The main feature I’m interested in is nuclear intensity of the GFP signal over time for each individual cell that remains in all of the images. I’m not sure what you mean when you say to discard the tracking information from 3).

thank you,
Mark