Characterizing non-uniform particle size

imagej

#1

The discussions here during the past week were welcome information for me. Analyze Particles and Nucleus Counter in particular yet, I need to understand how these might help me automate efforts to assemble particle size distributions.

I work with nano-particles routinely. Presently, these are irregularly shaped, overlap and often lack well-defined closed boundaries so, distinguishing among adjacent or overlapping particles can be quite challenging. With this description in mind would you expect either of these tools to offer me much?


#2

Good day,

please post a typical (not best) original raw image in TIF- or PNG-format and explain exactly what you like to do and what the desired end-result should look like.

Regards

Herbie


#3

Herbie,

Busy reading on these issues so, I just made time to respond. It does appear to me that ImageJ has plenty of computational power to improve both my effort to identify individual particles, increase particle count with fewer samples and make all of this more efficient and robust. I am attaching three representatives *.tiff for you to comment on.

I uploaded “Img10” to provide indications of what particles are shaped like and their overlap characteristics to make these issues less ambiguous. The other two images are more likely to be encountered by searching on the grid (3 mm Formvar coated copper). However, if you take the ideas these three images provide and let your mind take them to their logical extent you will obtain the result shown below, albeit at lower magnification resulting from the larger size of the conglomerate.

In my reading since posting the initial question it appears these samples may challenge segmentation algorithms. I consider it a welcome opportunity if this has placed me into a situation requiring further development of ImageJ, the Trainable Weka Segmentation, etc. I am all in! This is what we live for…:wink:.

Edit - In “proofing” response after finalizing it I cannot “see” images despite no indication to the contrary when uploading these. What do I need to do to correct this? It appears my method of attaching images is fine - one of the (4) methods known to work or be regarded as acceptable.

Rick


#4

Good day Rick,

the images arrived savely and I can perfectly display them.

However, I’m quite unsure about what you call a particle.
If I may assume that image 10 shows around 12 particles, I see no chance for a reasonable automatic segmentation of such particles from images 7 and 4. If you can’t visually segment the conglomerate, which I can’t, then there is no chance for a machine (there are only very few exceptions).

Sorry for this judgement!

Regards

Herbie

PS:
Your images arrive as RGB-images but I can’t see any colour signal.
I assume they are achromatic


#5

Hello Herbie,

Many thanks for checking in on this issue for me. Yes, you are correct in your interpretation of the particle count in image 10. Clearly, after reviewing the set of images, this is not really what I am pursuing. As regards your judgement that a machine cannot help us in this regard, I will simply have to live with this despite my desire no wanting to. Life presents us with these situations too!

Thank you again,

Rick