Cube-cut: vertebral body segmentation in MRI-data through cubic-shaped divergences

Area of Science:

• Medical Imaging
• Computer Vision
• Graph Theory

Background:

• Accurate volumetric segmentation of vertebrae in Magnetic Resonance Imaging (MRI) is crucial for clinical diagnosis and surgical planning.
• Existing segmentation methods may lack efficiency or precision, necessitating advanced algorithmic approaches.

Purpose of the Study:

• To develop and evaluate a novel graph-based method, Cube-Cut, for volumetric segmentation of vertebrae in MRI data.
• To enable user-defined control over segmentation smoothness via a parameter Δ.

Main Methods:

• A graph-based approach using a cubic template for volumetric segmentation of vertebrae in MRI.
• The Cube-Cut algorithm constructs a directed graph (s-t-network) where nodes represent voxel subsets and edge weights denote affinities to the vertebra or background.
• Minimal s-t-cut is computed in polynomial time to achieve segmentation, with infinite weighted edges enforcing smoothness.

Main Results:

• The Cube-Cut algorithm achieved an average Dice Similarity Coefficient (DSC) of 81.33% in quantitative evaluations.
• The C++ implementation demonstrated a running time of less than one minute for segmentation.
• The method allows for user-defined deviation from a regular cube through a smoothness value Δ.

Conclusions:

• The proposed graph-based Cube-Cut algorithm offers an efficient and accurate solution for volumetric vertebrae segmentation in MRI.
• The method's performance, demonstrated by a high DSC and rapid execution, suggests its clinical applicability.
• User-adjustable smoothness provides flexibility in adapting the segmentation to varying image characteristics.