# Tips for semi-automatic segmentation of fetal brain MRI

**URL:** <https://discourse.slicer.org/t/tips-for-semi-automatic-segmentation-of-fetal-brain-mri/17020>\
**Category:** Support\
**Tags:** segmentation, dicom\
**Created:** [April 10, 2021, 3:04pm UTC](https://discourse.slicer.org/t/tips-for-semi-automatic-segmentation-of-fetal-brain-mri/17020 "2021-04-10T15:04:30Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![GMA](https://sea2.discourse-cdn.com/flex002/user_avatar/discourse.slicer.org/gma/32/4221_2.png) [@GMA](https://discourse.slicer.org/u/GMA)\
**Post date:** [April 10, 2021, 3:04pm UTC](https://discourse.slicer.org/t/tips-for-semi-automatic-segmentation-of-fetal-brain-mri/17020/1 "2021-04-10T15:04:31Z")

</div>

Hello to all.  
I’m trying to segment fetal MRI brains acquired with multiplanar T2-HASTE sequences (see an example below taken from Radiopaedia).

 ![image](https://us1.discourse-cdn.com/flex002/uploads/slicer/original/3X/b/3/b33945cca205f73a07caa6d3b41f37f0637afda4.jpeg)  
I’m currently having a harsh time with manual segmentation which is really time consuming since I have to segment every slice and avoid CSF spaces and ventricles.  
I tried with a seed based approach, also a threshold based one without success. I also tried using a registration based approach but the coregistered images are reformatted into a different plane and I can’t get access anymore to a standard axial plane.  
Does anyone have any tips? Should I rely to a manual segmentation approach?

Thank you!

GMA

---

<div class="post-metadata">

**Author:** ![lassoan](https://sea2.discourse-cdn.com/flex002/user_avatar/discourse.slicer.org/lassoan/32/13_2.png) [@lassoan](https://discourse.slicer.org/u/lassoan)\
**Post date:** [April 14, 2021, 5:20am UTC](https://discourse.slicer.org/t/tips-for-semi-automatic-segmentation-of-fetal-brain-mri/17020/2 "2021-04-14T05:20:30Z")

</div>

What exactly you would like to segment?

Seed based approach should work fine, maybe you could post a few screenshots that illustrate what problems you are running into.

If registration-based approach works better then you can use that. You can resample the image with image planes at arbitrary orientation (using Crop module or any of the resample modules).