Divergence-Free Fitting-Based Incompressible Deformation Quantification of Liver

Area of Science:

• Medical Imaging
• Computational Anatomy
• Biomedical Engineering

Background:

• The liver is an incompressible organ, maintaining constant volume during respiratory movements.
• Accurate quantification of liver deformation is crucial for effective liver tracking and medical interventions.
• Existing methods for enforcing incompressibility during deformation analysis are often time-consuming and can weaken the deformation estimation.

Purpose of the Study:

• To develop a rapid and accurate method for quantifying incompressible liver deformation using a divergence-free fitting-based registration approach.
• To accelerate the incompressible deformation quantification process through a novel deep learning framework (DLF).

Main Methods:

• A divergence-free fitting-based registration method was proposed, mapping deformation to velocity in a diffeomorphic space.
• Fast Fourier-based Hodge-Helmholtz decomposition was employed to obtain divergence-free, curl-free, and harmonic fields.
• A deep learning framework (DLF) was constructed, utilizing an encoder-decoder network trained on an incompressible respiratory motion model.

Main Results:

• The proposed registration method accurately quantified incompressible liver deformation, achieving a mean liver overlap ratio of 95.33%.
• The developed deep learning framework (DLF) demonstrated a significant acceleration, being nearly 15 times faster than existing methods.
• The DLF effectively learned appearance-velocity correlations at a patch scale for accelerated motion quantification.

Conclusions:

• The proposed divergence-free fitting-based registration method offers a rapid and accurate solution for quantifying incompressible liver deformation.
• The integration of a deep learning framework substantially enhances the efficiency of liver motion quantification, making it clinically more viable.
• This approach provides a robust tool for liver tracking and related medical applications requiring precise deformation analysis.