My requirement is to find the volume of the sample within the MRI image which consists of the sample, water and air. I have the data in the form of intensities in a 3D array of size 128*128*128 in which the sample corresponds to black pixels and the water volume corresponds to lighter pixels. Also, the outside air volume and the air bubbles present which are black. I am trying to use watershed function to find the connected sequence of pixels and finding the one with maximum frequency (I assume the sample to be the largest connected sequence of dark pixels), but somehow it is not detecting the sample volume as connected one. Also, bwconncomp and the labelmatrix are not of much help. Any suggestions/ideas on how to find this ?
There are a couple of other ways to go about this. In particular, I think you'll know at least the approximate location and signal level of your sample. This suggests using a gray scale to binary cutoff to form an initial mask, then a region growing algorithm with an initial seed inside the known sample region to segment more exactly. Depending on how important it is to avoid boundary effects, etc, one can erode/grow the resulting mask.
I have a series of functions I used for this in 2 and 3 D, but would need to dig them out of my codes.
r/
David
1. Find all connected clusters of voxels in a certain intensity range
2. Merge clusters that belong to your sample (the exact strategy would depend on your sample shape and location characteristics). You can do it manually as first approximation and then see if cluster merging can be automated.
- Badges
- Science topic
- Similar topics
- Mathematics
- Algebra
- Matrix