# Segment comparison with transformed structure

**URL:** <https://discourse.slicer.org/t/segment-comparison-with-transformed-structure/11033>\
**Category:** Support\
**Tags:** segmentation\
**Created:** [April 7, 2020, 10:41pm UTC](https://discourse.slicer.org/t/segment-comparison-with-transformed-structure/11033 "2020-04-07T22:41:16Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![izzatizulkarnain](https://avatars.discourse-cdn.com/v4/letter/i/c68b51/32.png) [@izzatizulkarnain](https://discourse.slicer.org/u/izzatizulkarnain)\
**Post date:** [April 7, 2020, 10:41pm UTC](https://discourse.slicer.org/t/segment-comparison-with-transformed-structure/11033/1 "2020-04-07T22:41:16Z")

</div>

Hi, I am trying to get the dice coefficient and hausdorff distance to compare 2 structures from a TPS (red) and a DIR software (blue). I registered the blue GTV to the red GTV using the Transforms module and calculated the dice coefficient and HD. Despite the overlap of the two structures, why is the dice coefficient very small? Is there a better way for me to do the structure registration?

 ![Segment comparison](https://us1.discourse-cdn.com/flex002/uploads/slicer/original/3X/9/5/95a0ea3573733748234bc8b9ec9aab9f8f24bc7b.jpeg)

---

<div class="post-metadata">

**Author:** ![gregsharp.geo](https://avatars.discourse-cdn.com/v4/letter/g/6bbea6/32.png) [@gregsharp.geo](https://discourse.slicer.org/u/gregsharp.geo)\
**Post date:** [April 7, 2020, 11:03pm UTC](https://discourse.slicer.org/t/segment-comparison-with-transformed-structure/11033/2 "2020-04-07T23:03:19Z")

</div>

Something looks wrong. Can you try hardening the transform? That shouldn’t be necessary but worth a try.