|
|
Hello all,
I am trying to use ImageJ to analyze the connectivity of mitochondrial networks. I'm culturing C2C12 myocytes and imaging them via confocal microscopy. My cells are stained in a single channel (MitoTracker Deep Red), and I am only interested in the morphology of the mitochondrial network.
For my analysis, I follow a protocol suggested by De Vos and Sheetz (2007). The steps (somewhat modified) are as follows:
(1) Binarization via automatic threshold
(2) Skeletonization
(3) Gabriel Landini's BinaryConnectivity plugin
(4) Analyze histogram
The entire process is contingent upon successful and accurate binarization. Unfortunately, the cells in my images are somewhat heterogeneous in terms of staining intensity. That is, either the cells are not in the same plane (some are out of focus, and thus show weaker and blurred signal), or the cells do not take up the dye equally well. Global thresholding ("Make Binary") fragments the dimmer cells in the image such that they no longer accurately represent the original phenotype. A local threshold would be ideal. However, after looking through older posts in this forum, I found that "Dynamic Threshold 1d" introduces unacceptable amounts of noise and causes my mitochondrial tubules to split into lines. Perhaps I'm using it wrong?
Alternatively, I've tried cropping out individual cells to do the analysis on, in hopes of making the "global" threshold more local. However, for very long cells, this is not sufficient to prevent improper thresholding (and global thresholding may be more effective). And that's as far as I've been able to go on the issue. Suggestions of a free local/dynamic/"adaptive" thresholding plugin or procedure would be greatly appreciated. I have very limited experience with both image analysis and software writing, so I appreciate your patience as well.
Thank you for your help, and for reading this!
|
Locked
