Contour-Based Medical Image Segmentation with Implicit Anatomical Correspondence Learning

This demo showcases the HybridGNet model for contour-based medical image segmentation across multiple imaging modalities.

Instructions:

  1. Choose the imaging modality you want to work with
  2. Upload an image in PNG or JPEG format, or select an example from the dropdown
  3. Click "Segment Image" to perform automated segmentation

Note: Image preprocessing is handled automatically and will be reversed after segmentation to provide results in the original image coordinates.
Select Image Modality
Example Images

Example Image Sources:

All example images were obtained from Wikimedia Commons under open licenses. None of these come from the training nor testing datasets of the models.

Chest X-Ray Images:

Cardiac Ultrasound Images: Creative Commons Attribution-Share Alike 4.0 International. Source: https://commons.wikimedia.org/wiki/File:ProapsZCMiCh_.gif Images were extracted from the frames of the GIF animation.

Cardiac MRI Images: Public Domain. Source: https://commons.wikimedia.org/wiki/File:Multslice_short_axis.gif Images were extracted from the frames of the GIF animation.

Prenatal Ultrasound Images: Creative Commons Attribution-Share Alike 3.0 Unported license. Sources:

Author: Nicolás Gaggion
Website: ngaggion.github.io