search

tischi / fiji-plugin-morphometry

0 stars 1 forks source link

Spindle: Tubulin threshold #75

Closed tischi closed 5 years ago

tischi commented 5 years ago

@TobiasKletter @n-dempe

In the left image there is no cytoplasmic tubulin staining at all. Also the signal to noise is quite low in the left image. Do you know why that is?

image
TobiasKletter commented 5 years ago

@tischi In the left image there is no cytoplasmic signal because tubulin in these cells is not tagged with GFP but stained with SiR-tubulin dye (it only stains polymerized tubulin). Yes, the signal-to-noise is quite low in this data set. We have another data set using the same cells and staining technique, but with better signal, if you need more samples?

tischi commented 5 years ago

I got it working also with this data set. I sent Nils an e-mail to have a phone call to decide how to proceed, but he did not answer yet...

TobiasKletter commented 5 years ago

Some feedback for the new tubulin thresholding approach (in version 0.5.6):

I manually set the threshold for spindle segmentation in a couple of spindle images (three times blindly for each spindle and then took the average) and compared it to the measurements of Version 0.5.5 and 0.5.6:

20190821_Spindle_threshold_055_056

Is this helpful at all? I really don't know how well these manual measurements represent some sort of "ground truth" because it's of course a bit arbitrary when judging the threshold by eye... For some spindles, the standard deviation of the three measurements was even larger than 2 gray values (8bit).... Maybe we can try is out together on one or two images and discuss?

TobiasKletter commented 5 years ago

Idea: Get intensities from voxels surrounding DNA mask (exclude the ones expected to be inside the spindle = based on DNA lateral extent measurement)

TobiasKletter commented 5 years ago

This is now happening quite frequently:

HighZoom--W0000--P0001-T0175 tif-out HighZoom--W0000--P0001-T0188--0001 tif-out HighZoom--W0000--P0001-T0275--0001 tif-out

tischi commented 5 years ago

Could you please give me an example data set?

TobiasKletter commented 5 years ago

@tischi some milder cases, but still visible....will also look for more grave ones:

https://www.dropbox.com/s/g9203e425d5pq9i/SpindleVolume_HighZoom--W0000--P0001-T0285--0002.tif?dl=0

https://www.dropbox.com/s/1ips6dopm5ifr93/SpindleVolume_HighZoom--W0000--P0001-T0321.tif?dl=0

https://www.dropbox.com/s/rvfmqxp2w0p84kb/SpindleVolume_HighZoom--W0000--P0002-T0030--0002.tif?dl=0

tischi commented 5 years ago

I am working on this today. So if you have even better examples, would be good to share them.

tischi commented 5 years ago

In one of the images one can see the effect, where the tubulin is excluded by the DNA:

image
TobiasKletter commented 5 years ago

I am working on this today. So if you have even better examples, would be good to share them.

Here are some examples: https://www.dropbox.com/s/9icadjmmf5szdce/Spindle_mask.zip?dl=0

tischi commented 5 years ago

Ok. I think I found a better way: I am taking the average of the intensity in the cytoplasm and in the inside of the spindle (DNA radius - 2 micrometer). Like this it seems to work for the examples that you sent. I will upload the new version later today.

TobiasKletter commented 5 years ago

Seems to work great now (checked 30+ spindles)!

tischi commented 5 years ago

I'll close this issue for now. We can reopen if needed.