Hello,
For a DCE-MRI study, I would like to extract the signal intensity, through time, for a dozen organs and regions of the upper body.
I have successfully built a time sequence using my 40 volumes in time, drawn a segmentation mask for a single ROI, and experimented with the automatic segmentation with the excellent tutorial provided by @lassoan at Automatic segmentation of sequences in time , successfully propagating the registration through time.
This works with some success for bigger structures, but finer ones likes arteries and veins occasionally suffer from very bad registration; registration translates my veins into other organs for a few volumes, yet works fine for most of them…
I’m inclined to think that drawing more regions would add physiologically sound constraints on the registration, thus avoiding a transformation that misplaces, for example, the liver outside of the body, but this is purely speculative.
Finally, here is my questions:
- How would one solve this specific problem?
- Should each organ be separately segmented or all should be done simultaneously as I suspect?
- If registration of ROI works on most volumes, is it possible to manually adjust the few that do not align well?
- Is there a hidden functionality that I’m not aware of for this specific problem?
Looking forward to reading your ideas! Thank you for your help