Group Abstract Group Abstract

Message Boards Message Boards

Extract nuclei from images with dark cytoplasm?

Sotirios Maipas
Posted 7 years ago

Hello and thank you in advance for your help. Please, see, for instance, the following two images.

Cell 1 Cell2

Any ideas for a correct isolation of the nuclei? Unfortunately, these images are from our archive and they cannot be recaptured. The sample also cannot be re-prepared. This is a “tricky” challenge. Nothing seems to work for a correct isolation.

POSTED BY: Sotirios Maipas
3 Replies
Sort By:

Hi,

I recommend to apply a machine learning algorithm. In a first step you make the correct annotations on your own in a training sample. This means you draw the line for the nuclei.

In a second step you use this training sample and apply U-net as it is described here : https://community.wolfram.com/groups/-/m/t/1341081

if the images are not too large (say smaller than 192x192 pixels), I can help you and apply U-net for you.

Best Wolfgang
POSTED BY: Wolfgang Hitzl

Hi Sotirios,

here comes a simple first guess:

ClearAll["Global`*"]
web = {"http://community.wolfram.com//c/portal/getImageAttachment?\
filename=cell-4-9-1.jpg&userId=1438211", 
   "http://community.wolfram.com//c/portal/getImageAttachment?\
filename=cell-4-9-2.jpg&userId=1438211"};
imgs0 = Import /@ web;
thrh = FindThreshold /@ imgs0;
nucs = ColorNegate /@ MapThread[MorphologicalBinarize, {imgs0, thrh}];
MapThread[HighlightImage, {imgs0, nucs}]

This gives:

enter image description here

Does that help?

Regards -- Henrik
POSTED BY: Henrik Schachner

Dear Henrik,

Thank you very much for your very valuable help!!!

I will inform you about the results.

Thank you once again!

Best regards, Sotirios
POSTED BY: Sotirios Maipas
Reply to this discussion
Community posts can be styled and formatted using the Markdown syntax.
Tag limit exceeded
Note: Only the first five people you tag will receive an email notification; the other tagged names will appear as links to their profiles.
Reply Preview
Attachments
Remove
or Discard