Using AI Segmentation
Segmentation uses AI to automatically identify and highlight organs, structures, and lesions on your images. The results appear as colour-coded overlays on top of the image, making it easy to see exactly where each structure is.
Note: Segmentation requires a valid licence key.
Setting Up Segmentation
Option 1: Automatic Local Setup (Recommended)
This is the easiest way to get started. Kantrad handles everything for you:
- Open Settings, then go to Segmentation.
- Select Local (Python).
- Click Set Up Segmentation Server.
- Kantrad will automatically download and install everything it needs. This may take a few minutes the first time, as it needs to download about 1.3 GB of data.
- When the setup is complete, the server starts automatically.
After the first time, the segmentation server starts on its own whenever you open Kantrad.
Option 2: Connect to a Remote Server
If your department already has a segmentation server running, you can connect to it:
- Open Settings, then go to Segmentation.
- Select Remote Server.
- Enter the server address (ask your IT department for this).
- If the server requires a password or token, enter it in the authentication field.
- Click Test Connection to make sure it works.
- Click Save.
Running Segmentation on an Image
From the Viewer
- Click the brain icon in the viewer toolbar (in the second row of buttons).
- A panel opens showing which segmentation models are available.
- Choose a model and click Run Segmentation.
- After a moment, coloured overlays appear on your image showing the identified structures.
From the AI Chat
- Click the Segment button (shown in purple) above the chat input.
- Kantrad runs segmentation and shows the results in the chat window.
- The AI assistant automatically explains what was found.
- You can ask follow-up questions about the highlighted structures.
Adjusting the Overlay
Once segmentation results are showing on your image, you can:
- Change the transparency — Use the opacity slider to make the overlay more or less see-through (from fully transparent to fully opaque).
- See what each colour means — Check the label legend to know which colour represents which structure.
- Remove the overlay — Click Clear to remove the segmentation overlay from the image.
How You Know It Is Running
When the segmentation server is active, a purple Segmentation label appears in the top bar of the application.
What You Need
- About 1.5 GB of free disk space for the software and models.
- About 1—2 GB of available memory while the server is running.
- A graphics card (optional) speeds things up significantly, but it works without one too — just more slowly.