Measuring dendritic length in ImageJ

Hello there,

I’m new to the forum. I have been using ImageJ to analyse my data of Purkinje cell dendritic arbor. I was measuring the length of dendrites in binary and skeletonised image. I started using it yesterday after almost a year and figured out that the number of dendritic length in new measurements are completely different than earlier measurements. I couldn’t understand what went wrong and which parameters I put wrong or right there. I spent hours in figuring it out without any success.
Can someone help me here?
Here is how I did it?
Open the image =>make it binary=>skeletonise it=>measure

Thanks in advance!!

Hey @pradeep,
maybe is a simple thing, are you sure that have you calibrated the image that you want to measure?
Have a nice day,
Emanuele

Hi @pradeep,
Some ideas to try:
Is the scale factor set correctly? (Analyze > Set Scale)
Are you using the same parameters in Analyze > Set Measurements?
When you Make Binary, are the dendrite pixels going to 255 and the background to 0?
Are you accidentally measuring the whole image instead of just the dendrite skeleton (do you get the identical area measurement on different images)?

If none of these works, then maybe you can post an image and the results you got previously and now?

Hope this helps,
Theresa

1 Like

Dear Theresa and Emanuelle,

Thanks for the reply and inputs!! I haven’t calibrated the image. I tried setting up the scale and measurements, but that didn’t work either. I’m not measuring the the whole image and using rectangle, circle and freehand tool to measure dendritic length. I have forgotten all my settings and parameters from earlier measurement and can not recall them either. Here are my measurements for control:
Earlier => Now=>

So, if I want to start completely from scratch, what settings, scale parameters would you recommend? It would be of great help.

Thanks in advance!!

Hi Pradeep,

From looking at the numbers (this is the Area column, right?), I think two possible reasons for such a big difference are

  1. the image scaling and
  2. whether the measurement is looking at the ROI you drew or only the skeleton.

To fix the scaling:
Set a scale factor for every image. Then the measurement units will be in microns (actually square microns, but your skeleton is 1 pixel wide so it’s equal to length). If you have a wrong scale factor, or no scale set, you will get a different answer even if the number of pixels measured is the same.

To measure only the dendrites:
In Analyze > Set Measurements, check Area, and Limit to Threshold.
The effect of “Limit to Threshold” is shown in the screenshot.
On the first results line, I left the box unchecked and did Measure. I got the total area in the red ROI.
On the second line, after I checked Limit to Threshold, I did Measure again. This time I got only the black pixels that are within the red ROI.

Hope this helps.

Hi Theresa,

Thank you so much!! It was really really a great help. I had forgotten to set Limit to Theshold and also selected perimeter. With your suggested settings, I tried previously analysed data and it’s giving me the same number. However, I haven’t Set a scale factor for earlier measurements because I thought that my pixel measurements are equal to length, right?

Anyway, everything works fine now.

Have a nice day!!

Pradeep

Hi Pradeep,

Glad it worked :smile:

As for the scale factor, measurements in pixel units are fine if all the images were captured at exactly the same final magnification. This is affected by the objective lens, any extra lenses on the microscope (camera couplers, Optovar or zoom lens), the camera, and the camera binning. If all those factors are the same, then the pixel size is the same in all images and you can compare measurements directly using pixel units.

Thinking ahead, though, you probably want to publish your results someday, and for that, you will need to put a scale bar on your figures. So at some point you should find out the pixel size in microns anyway (for a tutorial, follow the scale factor link in my previous message). After you do that, you can convert all your pixel measurements into microns, which will be more meaningful for your readers.

Just make sure that you do all the image measurements in pixels OR microns but don’t mix the two!

Best,
Theresa

Hi Theresa,

You are right. I’m actually doing it for the publication. My all images were taken on the same microscope and 20x mag. I’m converting pixels to micron differently than suggested by ImageJ. I had a measured 50 micrometer square area which is 18362 pixels. I have calculated it for 50 micron and then for 1 micron as 2.71 pixels.
This way, I’m calculating mean dendritic length, standrad deviation and standard error of mean. This will go as a supplementary data although the graph values will be area and length in pixels. Branchpoints is another parameter used in this study. But, we are using dendritic length for the first time.

Thank you for your help and suggestions!!

Pradeep

1 Like

Hi Pradeep and tswayne

I need to measure dendritic length and branch-points in damaged and undamaged (treated and untreated) neurons. I went through this forum and kind of got an idea about how to do it. But since, I have never done it before, can you please either share an exact protocol including what type of image I need - meaning brightfield images or ok or I need to have fluorescent image? And do I need confocal images? How to differential between axon and dendrite?

Also are there any other factor that can be measured to compare damaged and undamaged neurons?

Looking forward to hear back. Thank you very much in advance for your help.

Chandani

Hi Chandni,

Make sure that your all images were taken at same mag and having a same resolution (Mine were 1600x1200; 8 bits). It doesn’t matter if your images are fluorescent or brightfield. I believe that having brightfield images make your efforts a bit less. Open Fiji or ImageJ, set up your measurements: Analyze -> Set measurement. In set measurements: check “Area” -> Area Fraction -> Limit to threshold and save the settings. Now, go to file and open your image. If your image is fluorescent, then go to “Process” -> “Binary” -> “Make Binary”. Your image is inverted to Grayscale. Again follow the same protocol, “Process” -> “Binary” ->, but this time you select “skeletonize”. You get a skeleton of your dendritic arbor. Use freehand, square or oval tool under File- Edit- Image to cover up your dendritic arbor. Then, go to Analyze -> measure. You get the pixel value. That’s your dendritic length in “Area” column. For dendritic branchpoints, use Sholl analysis or do it in photoshop by using marker on dendritic branchpoints and count them manually.
I hope this is helpful enough.
Cheers!!
Pradeep

1 Like

P.S. Fluorescent images would be okay. Axon is generally single, long and starts beneath the cell soma and branched terminally to make synapse. However, dendrites are on the top of the cell soma, very branched, usually short and tapering and many in numbers. It depends also on neuronal type. For example, Granule cell has 3-4 simple and unbranched dendrites, but Purkinje cell has a complex, much branched and elaborate dendritic arbor.

Hi Pradeep,

Thank you very much for the detailed instructions. I will follow this method and see how it goes. I am using following neurons, https://www.sciencellonline.com/products-services/primary-cells/human-neurons.html. These are human cortical neurons.

Again, thank you for your response.

Chandani

Hi all,
just stumbled over this discussion and wanted to give my two cent opinion on it because initially developing a tool to measure the corrected length of objects like dendrites I ended up committing the same mistake using the area.

This will have one measurement inconsistency and one wrong assumption:

  1. The area measures on a binary skeleton does not reflect the objects length because created by erosion (if I am not incorrect) and therefore all skeletons will always be shorter than the actual object. This error will be consistent if your original objects have all the very same width but will become bigger with increasing object width at the end points due to stronger shortening during skeletonization.

  2. The area of a skeleton unfortunately does not reflect its length. This would only be true if the line is perfectly horizontal or vertical. But since you will have parts of the line consisting of diagonal pixel distribution the length at this part is actually the sqrt(2) (square root of 2 ). And this finally will also have a certain inaccuracy. But at some point we might neglect it.

You also might have a look at the former discussion on this topic which also provide a comparison of different methods of length measurements and the differenced between them.

Furthermore, potentially the links to tools in analyzing e.g. dendritic length:

  1. Morphology tools by @gabriel

  2. Simple neurite tracer (►Plugins►Segmentation►Simple neurite tracer) by Mark Longair and @tferr

  3. Skeletonize (2D/3D) (►Plugins►Skeleton►Skeletonize(2D/3D)) by @iarganda

4 Likes

Thank you, @biovoxxel, for the correction and the links!