Lamellar thickness

I’d like to know if exist the possibility to measure the thickness of the lamellar structure (see the image). The images show an alternating sequence of dark and light lines where the dark regions correspond to the stained electron dense material. We attribute such images as being due to stacks of crystal lamellae where the white bands correspond to thecrystal core, which remains unstained and the dark bands to the disordered or amorphous lamellar surface and interlamellar material. So I’d like to measure the thickness without segmented or freehand line on the image. I’d like to use a procedure that exclude if possible the manual direct mesure on the image (that’s could introduce some error of interpretation). Many thx for your help.

Original tif image (click to download)



Hey @marco

I was unable to access your image… would you mind uploading it again?

eta :slight_smile:

strange because I’m able to read that. I could send u as attached file if u give me your e.mail.


I am mostly worried whether or not others can also access your dataset. It may not just be me…

But in the end - I was able to access your image using a different browser.


This is a known issue with tif images: the forum allows to embed them, but some browsers don’t show them. You can however always right-click and download the original image (that’s why wouldn’t like to disable tif upload in general…).

When I see those posts, I usually edit them, embed a png version of the image, and include a link to the original tif image below. Here’s what I did to edit the post above:

  • In your browser, right-click on the broken tif image and copy the link location
  • In ImageJ, use File > Import > URL… to open the image
  • Still in ImageJ, use Edit > Copy to System
  • Edit the forum post, paste the copied image (it will be in png format) and include the tif link in an extra line :slight_smile:

In order to achieve an unbiased (automatic) measurement, you should first segment your image to label the objects of interest.

Have a look at the Trainable Weka Segmentation plugin. There are quite some forum topics discussing this, e.g. those tagged with machine-learning.

Hi Jan,
thx for your suggestion but I’m not so expert in that topic. Could u apply in my img the algorithm suggested by u ?Just for understand as the result should be.


So my suggestion would be along these lines:

  1. Run Trainable Weka Segmentation to train a classifier on whatever an experienced scientist (i.e. you) considers to be crystal lamellae:

    FYI, I used the following settings:
  2. Create the result segmentation, threshold it and make binary:
  3. Use for example the Local Thickness plugin to measure the thickness. Here’s the result displayed with a false-color LUT:

You’ll find documentation on all those steps on the ImageJ wiki as linked above.


Hi Jan,
I’ll try your procedure and I’ll correspond u the feedback of the result. Many thx for your help.

Hi Marco,

I agree weka is a nice way to do it, but in my experience it requires quite a few images for training and can end up taking longer than manual analysis, especially if this isn’t a frequent assay.

An alternative would be to manually measure using random sampling. You can do this by applying a grid to each image in imageJ and only counting thickness of structures at the grid intersections, for example. Obviously, in an ideal world you would have the analysis blinded by another person doing it, but very few of us have access to such precious resources. As a compromise, you can randomly rename the files with the filename randomiser plugin. This can work nicely, as long as you haven’t managed to memorise what your images look like!

Apologies if this reply is too late for you.


1 Like

Hi Glyn,
u r welcome and I’ve tried this approach in the past. It’s similar to the result obtained with the method that describes the procedure for determining the average measured chord length of the randomly truncated cells (e.g foam,cellular polymers). Many thx for your suggestions.