I have been trying to analyse neurons for spine density and size using ImageJ.The images are from fixed brain section taken using confocal microscope. but the background fluorescence is interfering with the analysis as the software is unable to distinguish between actual spine and background noise. What plugins/macros can be used to reduce noise and clearly outline the structure of the neuron. Please do not hesitate to ask for more details in case you think you could help me out ! Please find the iimage attached. Many Thanks, - Raj
In this case, the simplest thing you can try is a median filter.
Then if precision is what you need, give NeuronJ or Simple Neurite Tracer a try. The plugin helps you semi-automatically trace all the dentrites to end up with a clean segmentation which you can then analyse much more easily. Simple Neurite Tracer works with stacks and is quite well documented.
The semi-automated part makes it slow, but in cases like these, nothing beats human eyes.
http://fiji.sc/Simple_Neurite_Tracer
http://www.imagescience.org/meijering/software/neuronj/
Anisotropic diffusion is proven to be a good image filtering technique that deals with noise while preserving objects' outlines (edges). Gabor-based and Hessian-based filters are used to highlight structures in retinal fundus images and to segment blood vessels which could be something handy for your work.
Thanks a lot Olivier, for the link to Neurite tracer. But do you know if I could extend that analysis to identify dendritic spines on the segment and measure their shape and size?
Many Thanks Abbas... but did you say 'Anisotropic diffusion' to filter the noise from the microscopic images ?
Yes Raj, it is also called Perona-Malik algorithm, it works much better than Gaussian filtering according to some resources (see Nixon & Aguado, "Feature Extraction & Image Processing for Computer Vision", Ch. 3, pp:114-120.
Check also the other two filters I mentioned earlier, they are worth trying for the example image you gave (after pre-processing of course).
Thank you very much, Abbas ! Thats quite a great help. Much appreciated !
Hi Raj,
Some people have alreadz asked this for imageJ apparently. Perhaps you should contact the authors and see if they have something in that direction. However from the image you submitted, no dentritic spines are visible, so I cannot tell you wether the imaging conditions are optimal for your needs in terms of resolution.
https://www.researchgate.net/post/Does_anyone_have_information_on_dendritic_spine_shape_analysis
Raj did you find a solution that was optimal for this problem?
Hello everyone, we just published a preprint on dendritic spine analysis. Check : https://www.biorxiv.org/content/10.1101/2020.09.12.294546v2 and https://github.com/argunsah/spines . Contact me if you need help :)
