Segmentation and 3D Visualization of Spinal Motion Segments from MSCT Images Using a 3D U‑Net Framework

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Spinal Surgery

Background:

• Accurate segmentation of spinal motion segments from MSCT images is crucial for clinical evaluation and surgical planning.
• Existing methods may lack the automation and precision required for efficient clinical workflows.

Purpose of the Study:

• To develop and validate a fully automated framework for segmenting spinal motion segments from MSCT data.
• To enable precise 3D reconstruction and visualization of patient-specific spinal models for surgical planning.

Main Methods:

• A framework combining a 3D U-Net for segmentation, a watershed algorithm for vertebra separation, and a marching cubes pipeline for 3D reconstruction.
• Development and testing on a dataset of anonymized lumbar-spine MSCT scans from 50 male patients.
• Performance comparison against Dense-U-Net and nnU-Net baselines.

Main Results:

• The proposed 3D U-Net achieved high segmentation accuracy (Dice: 0.96, Jaccard: 0.93).
• Vertebra separation demonstrated a 95.8% success rate with minimal boundary deviation (1.9 mm).
• The system outperformed baseline models and included a GUI for interactive model manipulation.

Conclusions:

• The developed framework offers a reliable and clinically practical solution for MSCT-based spinal motion segment analysis and 3D visualization.
• This technology supports improved surgical planning and has potential for integration into orthopedic workflows.
• Future research will focus on multi-center datasets and dynamic (4D) motion analysis.