Real-time biomechanical modelling of the liver using LightGBM model

Area of Science:

• Biomedical Engineering
• Computational Mechanics
• Surgical Simulation

Background:

• Accurate, real-time biomechanical modeling of the liver is crucial for computer-assisted surgery.
• Finite element method (FEM) offers high accuracy but suffers from high computation costs, limiting real-time applications like virtual surgery.
• Existing methods struggle to balance accuracy and speed for dynamic organ modeling.

Purpose of the Study:

• To develop a computationally efficient method for real-time biomechanical liver modeling.
• To enable accurate prediction of liver deformation under various forces for surgical simulations.
• To overcome the limitations of traditional finite element methods in speed-critical applications.

Main Methods:

• A liver model with realistic biomechanical properties was created using FEM.
• A dataset of liver deformations was generated by applying forces (0.1–0.5 N) to different surface points.
• A tree-based LightGBM regression model was trained on this dataset for real-time simulation.

Main Results:

• The LightGBM model achieved high accuracy, with mean absolute errors (MAE) of 0.0774 mm (x-axis), 0.0786 mm (y-axis), and 0.0801 mm (z-axis).
• Root mean square errors (RMSE) were 0.0591 mm (x), 0.0609 mm (y), and 0.0622 mm (z).
• The LightGBM model computed deformations in 33 ms, significantly faster than FEM (29.91 s).

Conclusions:

• The LightGBM model provides a highly accurate and computationally efficient solution for real-time liver deformation simulation.
• This method lays the groundwork for developing advanced real-time virtual surgery systems.
• The findings are particularly relevant for simulating liver deformation during minimally invasive surgeries.