Abstract:
Predicting soft tissue deformation after maxillofacial surgery is a critical, yet challenging, aspect of surgical planning and outcomes assessment. Traditional finite element models (FEMs), while accurate in capturing the complex biomechanical behavior of tissues, often suffer from prohibitively long computation times, making them impractical for real-time clinical use. This paper addresses this challenge by proposing an innovative nonlinear regression model, specifically Incremental Kernel Ridge Regression (IKRR), for the rapid and accurate prediction of soft tissue deformations following maxillofacial surgical procedures.
The core of the proposed biomechanical approach involves extracting features from a pre-computed Finite Element Model (FEM), which serve as inputs for the regression model. This leverages the detailed biomechanical insights provided by FEMs without incurring their full computational cost during actual predictions. A significant advancement introduced is the incremental learning capability, which allows the model to continuously improve its accuracy as more patient data becomes available without needing to retrain from scratch. This is particularly advantageous in a clinical setting where new patient cases are constantly encountered.
The IKRR method was rigorously validated using both synthetic data generated from biomechanical simulations and real patient data following orthognathic surgery. The results demonstrate that the IKRR model achieves comparable prediction accuracy to traditional FEMs while drastically reducing the computational time, making real-time prediction feasible. Specifically, the average error for IKRR was found to be lower than traditional FEM and Kernel Ridge Regression methods. This advancement in computational biomechanics provides a robust and efficient tool for clinicians to accurately predict soft tissue changes, thereby enhancing surgical planning, improving patient communication regarding expected outcomes, and refining the overall efficacy of maxillofacial surgeries.
