Comparative Analysis of Non-Rigid Registration Techniques for Liver Surface Registration

Area of Science:

• Medical Imaging
• Surgical Navigation
• Computational Anatomy

Background:

• Non-rigid surface-based soft tissue registration is vital for surgical navigation but faces challenges with complex intra-operative data.
• Accurate registration enables real-time visualization of patient anatomy, enhancing surgical guidance.
• Existing registration methods, especially for liver applications, often lack accessibility.

Purpose of the Study:

• To comparatively analyze open-source, non-rigid surface-based liver registration algorithms.
• To identify strengths and weaknesses of different registration methods.
• To determine an optimal algorithm for pre- to intra-operative liver registration.

Main Methods:

• Evaluated four non-rigid registration algorithms: three optimization-based and one data-driven.
• Assessed algorithm robustness against reduced surface visibility and increasing deformation levels.
• Quantified registration accuracy using root mean square error (RMSE) between pre- and intra-operative liver surfaces.

Main Results:

• The Gaussian Mixture Model - Finite Element Model (GMM-FEM) method demonstrated consistently lower post-registration error.
• GMM-FEM outperformed other methods under conditions of reduced visibility and increased deformation.
• All tested algorithms showed increased error with higher deformation and reduced visibility.

Conclusions:

• The GMM-FEM method presents a robust and promising solution for non-rigid liver surface registration.
• This approach can improve accuracy in surgical navigation systems integrating pre- and intra-operative data.
• Open-source comparative analyses are crucial for advancing surgical guidance technologies.