The training images for the label free prediction

Software & Code Miscellaneous
1

Hi, sorry for this maybe very basic question. I’m very new to image processing.

I downloaded the label free tom20 images from the website: https://www.allencell.org/data-downloading.html#sectionLabelFreeTrainingData

When I used aicsimageio package to read the czi files, I found that each image has 7 channels. [‘CMDRP_1’, ‘CMDRP’, ‘EGFP_1’, ‘EGFP’, ‘H3342_1’, ‘H3342’, ‘Bright_100X’]
and the metadata for a specific image is:
<Element ImageDocument at 0x12de47eb0>

My question is I can’t find specific information about what each channel means.
I know that EGFP stains for the mitochondria, H3342 for DNA and CMDRP for membrane. But I don’t know what’s the difference with EGFP and EGFP_1.

After looking into actual pixel values of each channel, my assumption is that EGFP_1 is the channel for the stain, and EGFP is the binary segmentation channel with cell numbers.

Is there anyway I can find documentation or someone to confirm this for me?

Thank you so much!!

2

Hey Hanxix!

All _1 channels i.e. EGFP_1 are empty channels collected to decrease noise introduced by the filter wheel changing position at each plane (each plane was acquired with interwoven channels). These are needed in the training set in order to keep constant the time between z-plane (training set vs data set used for prediction need to be acquired with the same timing between Z-planes and between channels).