I'm trying to analyse images of animal tissues produced with a 3DHistech slide scanner. I usually use ImageJ but have problems fitting large enough images in the memory. Are there any other free alternatives, or tricks that could be used with ImageJ? My analysis is simple, just detecting & measuring circular holes of certain size (adipocytes). The images are like 25000 pixels x 25000 pixels.
How about Matlab? Its pretty good with images and you can easily find some ready made functions for a job like that (first google result: http://blogs.mathworks.com/pick/2008/05/23/detecting-circles-in-an-image/ ).
vaa3d is a free 3d image analysis software which can handle large data set. but i have not done any analysis part in it so can't say measurement of holes in it. user interface is too simple to give a try and it is worth trying :)
You just need a powerful computer. ImageJ can run the Large TIF(TIFF) files but needs a lot of RAM. Extract the files in like a 50% resolution or even 25% TIFF and load them to a computer which is at least 4 cored and has 64 GB of RAM with ImageJ installed. Should work! I do the same trick with my Zeiss scans.
Just be prepared for quite a time consuming jobs. Image analysis is proportional to computer power.
Best,
Ivan
If you don't have a computer with sufficient RAM then my suggestion is to process the images in a tile-based manner but be sure to have sufficient overlap between the tiles in order not to miss anything.
As far as I know, remote sensing image analysis software, e.g., ENVI, ERDAS, PCI can handle very large images.
Yes, Size of image and memory of analyzing device correlates. MATLAB comes with inbuilt functions which makes analysis or algorithm development easy. Imagepro is also available for image analysis and LABVIEW also has option of image processing.
May you will have a look
http://pdl.perl.org/
related to your problem:
http://nbviewer.ipython.org/github/demianriccardi/p5-IPerl-Notebooks/blob/master/Misc/TLCFun.ipynb
PDL is desigened to work "inplace" on your data, keeping memory small, avoiding internal copies
If you are just detecting holes in the image, you can always divide you image into smaller regions and carry out the detection on individual region which would be much smaller an image. If you worry about the boundary, you can have a sliding window on the image that goes through all regions with the same smaller window size.
Please contact Marek Krótkiewicz my co-author in regard of possible use of APR. It is image analysis software using morphological analysis briefly described in one of our publications.
Chapter Morphological Analysis of Binary Scene in APR Integrated Environment
Hi, if you're still working in this area you could try QuPath - https://qupath.github.io
It might already be able to do what you need directly, but if not then it can be used to interactively extract regions from the image and send them to ImageJ for further processing at any resolution - and optionally receive any detected areas back again into QuPath. This should help avoid any memory issues arising from trying to work with ImageJ and whole slide images directly.
You can also write scripts or extensions for QuPath if you need to automate the process a bit more.
There is some more information about both these options in the QuPath documentation at https://github.com/qupath/qupath/wiki - including a section on how to use QuPath along with ImageJ.
(Sorry this answer is a bit late, although QuPath was only actually released open source within the last few days...)
Hi Mikael,
We designed Cytomine for very large (histology) images, it's free and open-source:
http://www.cytomine.be/
http://www.cytomine.be/
- Badges
- Science topic
- Similar topics
- Correction