Example Problem: Help to count nuclei in microscopy images

Hi,

I have several hundred images looking like this:

.nucleus_image

They show nuclei imaged with a microscope and I need to count how many nuclei (= bright spots) are in there. I don’t have time to do this by hand right now.

I have access to a Windows laptop at work and could use this, but I don’t really know how to tackle a problem like this. I can also upload all these images somewhere.

Important: This is an example question to give an idea what kind of questions can be asked about microscopy images and how the follow up interactions could look. The problem at hand is based on the 2018 Kaggle challenge for nuclei detection https://www.kaggle.com/c/data-science-bowl-2018. For the problem at hand, the most easy to use tool is probably cellprofiler, as pointed out by @lesolorzanov.

2 Likes

Example Answer

Hi @cpape, perhaps this is doable with ilastik, which is a well known tool for image segmentation and supports cell counting.

Do you expect this problem to be recurring? If so, I can walk you through using ilastik for your problem via e.g. Skype or Zoom. If not, I can process the volume for you and upload the results.

2 Likes

Dear @nasim.rahaman,

thank you for pointing out that tool, it looks promising.

Maybe in the future, but for now it would be enough if the images we have are processed.

I have uploaded the images we have now:
https://github.com/mpicbg-csbd/stardist/releases/download/0.1.0/dsb2018.zip
Could you process those and send them to me or upload them somewhere?

Sure thing, I’m working on it. :+1:

1 Like

Hello! CellProfiler is the standard way these days todo this kind of job. They actually have already pipelines built you just load them and start counting. https://cellprofiler.org/examples/#human-cells let me know if you need help setting it up. Just download CellProfiler and the project and tell it where your images are.

1 Like

Hi @cpape,

the counts (per image) seem to be the following, as computed with ilastik:

127.670693161
18.3515386581
22.2584173679
87.4480013591
48.6618075371
78.7455368042
43.2528944016
14.7031588554
228.258145779
41.4835295677
55.0056428909
138.427466393
27.2136740685
91.5094432831
253.176972866
24.8782454729
74.7734298706
42.2979402542
30.4547252655
63.0923919678
19.9866583347
15.3438867331
44.5774774551
80.880537957
29.7715163231
56.2379493713
20.5476264954
106.463568211
11.5674357414
102.331080453
194.058986212
125.601499081
18.4643909931
52.0299139023
64.6797018051
64.2429695129
71.8264980316
372.045844078
13.9341617823
15.7366724014
48.8084907532
26.5235254467
393.677540213
154.44021225
32.9114780426
14.4799840152
72.9799461365
19.252840519
88.363196373
425.482092202
12.5645100474
414.970721006
69.2341632009
42.3954429626
24.0910334587
55.6314048767
89.2368620411
48.4514169693
38.4975566864
154.654726267
121.268774986
38.0044312477
400.167327046
119.52159214

Let us know if you need anything else.

1 Like

Dear @nasim.rahaman thank you, this solves our issue for now.
Is there any way we can rerun this on our own for new images?

1 Like

Hi @cpape, glad to help.

You can find a tutorial using ilastik here: https://www.ilastik.org/documentation/counting/counting

(See also @lesolorzanov’s link with CellProfiler)

If you’re stuck, please let me know and we can schedule a Skype call.

1 Like

Ok, thank you! This solves our problems for now.

Just a side note: questions of this form could also be posted on image.sc – a forum populated by many bioimage analysis experts… just in case… :wink:

3 Likes

You could use Stardist a Fiji plugin based on machine learning. It works accurately in 2D images (Also in 3D but is not implemented un Fiji).