Neurite length measurement


#1

Dear Cell Profiler Team

I am new to the software so please forgive me if my pipeline (attached) is naive.

I am trying to measure neurite length using cell profiler.

I have four stains: DAPI, neurofilament (647 wavelength), map2 (568, dendritic), SMI312 (488, supposedly axonal).

My pipeline doesn’t appear to be working at all and I have fiddled with it a lot now. I am at the moment concentrating on the “dendritic” read-out but ideally I would like to calculate lengths for dendritic, axonal and neurofilament.

My other problem is that the DAPI nuclear staining is not clean in my system, and there is a lot of debris not attached to a neurite that needs to be filtered out.

Any help would be much appreciated! I am attaching one czi file and my pipeline as it stands.

Many thanks indeed!


#2

testpipelineam.cpproj (1.0 MB)


#3

34d6 lmf 6jun-26.zip (4.5 MB)

Here is the czi file


#4

@Arpan_Mehta I tinkered with your pipeline a bit to at least have it output neuron measurements. Please try it out!

Here are a few observations:

  • The DAPI/DNA channel is very difficult to process. As you mentioned the signal is low and there is a lot of debris. I was able to segment the neurons without using this channel. If the DNA channel is essential then you could isolate the true DNA by masking with another channel, or, if you’re able, to obtain higher quality images.
  • The illumination is noticeably uneven, so illumination correction will improve segmentation results a great deal.
  • Since the DNA channel was poor, entire neurons were segmented directly on the panneuronal channel instead of trying to use segmentation of the nuclei as a seed.
  • When performing morph operations, like skeletonize, choose number of repeats to be “forever”. The skeletonized objects from this module can then be used in the MeasureNeurons module.

testpipelineam.cppipe (17.5 KB)


#5

Dear Kyle

This is extremely helpful - thank you for your time.

So am I correct in my understanding that the two neurite length measures outputted of interest (presumably already in micrometers - does the programme get this automatically from the metadata?) are:

Mean_NucleiShrink_Neuron_TotalNeuriteLength_denSkel
and
Median_NucleiShrink_Neuron_TotalNeuriteLength_denSkel

The data look reasonable, except for one case the media number is zero whereas the mean is a sensible figure (around 100). How might this be possible?

Thanks, again,
Arpan


#6
  1. The length is in the unit of pixels. This can be converted into a physical distance using the image metadata created by the microscope. A convenient way to view the metadata is with the FIJI software (Image > Show Info…).

  2. I’m not sure why the median would be zero. It seems like a neurite length should at least be of 1 pixel; otherwise what exactly was segmented? Are the number of objects in this image reasonable?