What are the current research areas and challenges in brain image segmentation?
There are no well defined segmentation algorithms because it's hard to define the borders of objects. A threshold might work in one case but not in all. This is a drawback.
Actually, there are many software programs that are designed to segment the brain based on T1-weighted (and/or T2-weighted) MR images. For example, FreeSurfer and FSL. I prefer FreeSurfer because it is fully automated from a single command and produces very accurate results. Version 6 also segments hippocampal subfields quite accurately. I have segmented thousands of brain MR scans with these software packages and only had problems with scans that had gross abnormalities (e.g. tumours).
My ebook provides step-by-step instructions: http://jeromemallershandbookofstructuralbrainmrianalysis.yolasite.com
Jerome
Satisfied with the answer of Ajay, There is no specific segmentation algorithm, that can work for all types of applications. However, you can try mean shift, graph cut algorithm, level Set method, active contour mode for image segmentation.
refer below article for challenges:
http://asrjetsjournal.org/index.php/American_Scientific_Journal/article/viewFile/2571/1020
One challenge is with the acquired data going into segmentation algorithms; usually resolution and tissue contrast are too low for most algorithms to accurately identify many small subcortical regions and subregions. However, for the future, UHF + MP2RAGE seem to point in the right direction. A multimodal approach (additionally using T2, SWI, DWI, etc scans) is probably best if one can obtain them for each subject.
- Badges
- Science topic
- Similar topics
- Comment