I'm looking for a program that will do automatic approximate image segmentation for prostate T2 MRI images.
The program should be:
The program does not need to be:
Sample Prostate MRI Image with segmentation attached
from http://appsrv.cse.cuhk.edu.hk/~lqyu/prostate-segmentation/
Jeffrey,
Automatic segmentation should be a relatively easy procedure that could be implemented on most systems, especially with you wanting only a very quick
approximation. I come from the remote sensing (earth and bottom of ocean imaging, including optical, sonar, and radar systems) community but a digital image is a digital image. Is there anyway you can make an example image available?
Pat
As Pat says above, an example image would help. Nevertheless, here is a link to a free segmentation method that is fast (based on the the histogram only) and does not need an initial seed or number of clusters.
https://se.mathworks.com/matlabcentral/fileexchange/37556-a-fast-image-segmentation-using-delaunay-triangulation-revisited
If you decide to use it let me know how it behaveson your data samples.
Regards,
Abbas
Abbas Cheddad and Pat S. Chavez, Jr.
Attached a sample image to the question.
Sample Prostate MRI Image with segmentation attached
from http://appsrv.cse.cuhk.edu.hk/~lqyu/prostate-segmentation
I will try the histogram method out. The problem I am having is the more elaborate convolutional network based approaches sometimes hang.
Ok, I realised that the method above is more suited for natural images.
Well, what about these two methods:
- multithresh (applies Otsu with multi-level thresholding)
- kmeans
Unfortunately, in both, you need to supply the number of clusters. You can estimate that from a rough examination of a number of samples.
Jeffrey,
Since I am not from the medical community (but am interested in applying what I have learn over 45 years working with digital images) could you perhaps explain a little what it is I am seeing and looking for? Do you have a single image representing the before and one representing the after? Can they be made available in TIF or JPEG formats? Sorry for the 'basic' question.
Pat
Jeffrey, there is no ready-to-go software that would automatically perform prostate MRI segmentation right now, maybe with the exception of radiological workstations (but still doubt it).
The link and image you've attached show the deep learning approach, which would require you to implement the neural network and run it. This needs significant computational power, proper setup and a trained model, all of which can be difficult. However, this IS the way to get closest thing to accurate and fully automated prostate segmentation.
We have developed an online segmentation tool with implemented AI-models, one of which can segment the prostate gland automatically. Find it here: www.medseg.ai.
A somewhat late answer but perhaps useful for future projects!
- Badges
- Science topic