Skip to main content

​
What is SpinEPS?

SpinEPS is an AI model that automatically segments spine structures from T2-weighted MRI scans. It identifies:
  • Vertebrae — C1 through S1 (cervical, thoracic, lumbar, sacral)
  • Intervertebral Discs — C2/3 through L5/S1
  • Spinal Canal — The central canal containing the spinal cord
3D segmentation result

​
How It Works

SpinEPS uses a two-phase approach based on nnU-Net:
  1. Semantic segmentation — Classifies each voxel into tissue types (vertebra, disc, canal, background)
  2. Instance segmentation — Separates individual vertebrae and discs (e.g., L1 vs L2)
The result is rendered as 3D meshes overlaid on the MRI volume.

​
Requirements

  • Spine Helper must be installed and running (see Helper Setup)
  • T2-weighted MRI is optimal (the model is trained on T2W data)
Segmentation is optimized for T2-weighted sagittal spine MRI. Results on other contrasts may vary.