# Multi-Class Segmentation with CT + Aligned STL (Dental Dataset)

**URL:** <https://discourse.slicer.org/t/multi-class-segmentation-with-ct-aligned-stl-dental-dataset/46636>\
**Category:** Support\
**Tags:** segmentation, dicom, 3d-model\
**Created:** [April 2, 2026, 6:37am UTC](https://discourse.slicer.org/t/multi-class-segmentation-with-ct-aligned-stl-dental-dataset/46636 "2026-04-02T06:37:36Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![abhijeet2410](https://sea2.discourse-cdn.com/flex002/user_avatar/discourse.slicer.org/abhijeet2410/32/11196_2.png) [@abhijeet2410](https://discourse.slicer.org/u/abhijeet2410)\
**Post date:** [April 2, 2026, 6:37am UTC](https://discourse.slicer.org/t/multi-class-segmentation-with-ct-aligned-stl-dental-dataset/46636/1 "2026-04-02T06:37:36Z")

</div>

Operating system: Window  
Slicer version: 3 D Slicer 5. 10.0

* * *

## 🧠 Question: Multi-Class Segmentation with CT + Aligned STL (Dental Dataset)

I am working on a **dental AI dataset pipeline** where I have:

- **CT scan (DICOM volume)**

- **Aligned STL models** (upper jaw, lower jaw, bite scans)

I am able to visualize everything correctly using VTK:

```auto
import vtk

def load_stl(path, color):
    reader = vtk.vtkSTLReader()
    reader.SetFileName(path)
    reader.Update()

    mapper = vtk.vtkPolyDataMapper()
    mapper.SetInputConnection(reader.GetOutputPort())

    actor = vtk.vtkActor()
    actor.SetMapper(mapper)
    actor.GetProperty().SetColor(color)

    return actor

renderer = vtk.vtkRenderer()

upper = load_stl("UpperJawScan.stl", (1,1,1))
lower = load_stl("LowerJawScan.stl", (0.9,0.9,0.9))

renderer.AddActor(upper)
renderer.AddActor(lower)

```

![image](https://us1.discourse-cdn.com/flex002/uploads/slicer/original/3X/5/1/516d457327ffed3a21e00d009430ab5b2fd30580.png)

![image](https://us1.discourse-cdn.com/flex002/uploads/slicer/original/3X/6/a/6a512656a8fbb43644285524c7721bb2b3ed3482.png)

So currently:

- CT volume and STL teeth are **spatially aligned**

- STL provides **clear tooth geometry**

- CT provides **voxel grid (needed for training)**

* * *

## 🎯 Goal : I want to create a **multi-class segmentation dataset** such as:

| Class ID | Structure |
| --- | --- |
| 0 | Background |
| 1 | Teeth |
| 2 | Gum |
| 3 | Impacted Teeth |

* * *

## ❓ Problem

- Teeth are **not clearly visible in CT slices** , but are **very clear in STL**

- STL is **surface mesh** , not voxel labels

- Exporting via `vtkOBJExporter` only gives geometry, not labels

* * *

## ❓ Core Question

**How can I generate a voxel-wise multi-class segmentation mask from aligned STL + CT data using tools like 3D Slicer or ITK-SNAP?**

 ![image](https://us1.discourse-cdn.com/flex002/uploads/slicer/original/3X/c/d/cd3deab46d2976934e65f20a854e8abae6d388d9.jpeg) This is **aligned STL + CT data. How to annotate it using 3D slicer ? White part show STL and transparent part show the CT**

* * *

## ⚠ Constraints

- Dataset is for **deep learning (3D segmentation / PointNet / UNet)**

- Prefer **semi-automatic or automated pipeline**

* * *

## 💡 What I’ve Tried

- Rendering CT + STL together → works

- Exporting meshes → no label info

- Manual annotation → difficult because CT does not clearly show teeth