# Accuracy evaluation

**URL:** <https://discourse.slicer.org/t/accuracy-evaluation/32460>\
**Category:** Support\
**Tags:** segmentation\
**Created:** [October 29, 2023, 6:53pm UTC](https://discourse.slicer.org/t/accuracy-evaluation/32460 "2023-10-29T18:53:36Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![Caterina](https://avatars.discourse-cdn.com/v4/letter/c/439d5e/32.png) [@Caterina](https://discourse.slicer.org/u/Caterina)\
**Post date:** [October 29, 2023, 6:53pm UTC](https://discourse.slicer.org/t/accuracy-evaluation/32460/1 "2023-10-29T18:53:36Z")

</div>

Dear support,  
I would like to know if there is a way to evaluate the accuracy of segmentation. I mean that I need a quantitative measure of how accurate the segmentation is.  
Thanks

---

<div class="post-metadata">

**Author:** ![pieper](https://sea2.discourse-cdn.com/flex002/user_avatar/discourse.slicer.org/pieper/32/8_2.png) [@pieper](https://discourse.slicer.org/u/pieper)\
**Post date:** [October 29, 2023, 7:25pm UTC](https://discourse.slicer.org/t/accuracy-evaluation/32460/2 "2023-10-29T19:25:06Z")

</div>

That entirely depends on what you are trying to measure and what kind of images you use to do the measurements. As a general rule, you would make a ground truth measurement of something like a phantom and compare it to what you can measure from images.

---

<div class="post-metadata">

**Author:** ![Caterina](https://avatars.discourse-cdn.com/v4/letter/c/439d5e/32.png) [@Caterina](https://discourse.slicer.org/u/Caterina)\
**Post date:** [October 29, 2023, 7:50pm UTC](https://discourse.slicer.org/t/accuracy-evaluation/32460/3 "2023-10-29T19:50:20Z")

</div>

Thanks for reply. In case of CT or rm dicom images and of anatomical regions segmentation, which is the best way to evaluate how accurate the segmentation is compared to real organs?

---

<div class="post-metadata">

**Author:** ![vet0282](https://sea2.discourse-cdn.com/flex002/user_avatar/discourse.slicer.org/vet0282/32/8747_2.png) [@vet0282](https://discourse.slicer.org/u/vet0282)\
**Post date:** [October 29, 2023, 8:09pm UTC](https://discourse.slicer.org/t/accuracy-evaluation/32460/4 "2023-10-29T20:09:19Z")

</div>

We have done some work on the accuracy of segmentation and 3D-printed  
replicas of bones.

Tow most common ways are

1. To directly measure geometry and compare distances between multiple  
points or landmarks.
2. To compare a segmented model derived from CT with a 3D surface scan of  
the model. So, this can be done in cadavers or dry bones.

---

<div class="post-metadata">

**Author:** ![Caterina](https://avatars.discourse-cdn.com/v4/letter/c/439d5e/32.png) [@Caterina](https://discourse.slicer.org/u/Caterina)\
**Post date:** [October 31, 2023, 12:55pm UTC](https://discourse.slicer.org/t/accuracy-evaluation/32460/5 "2023-10-31T12:55:45Z")

</div>

Thanks for reply. Can you please share with me the the work on the accuracy of segmentation and 3D-printed replicas of bones?  
Thanks