Abstract
Rib cage reconstruction is critical for maintaining chest rigidity, protecting intrathoracic organs, and preserving vital physiological functions. Although titanium has traditionally been used for reconstruction due to its mechanical strength and biocompatibility, its limitations have prompted the search for alternative materials. The finite element method (FEM) is widely used to assess implant performance through stress analysis, while advances in artificial intelligence (AI) now allow the integration of FEM with predictive modeling to efficiently estimate mechanical responses. This study aimed to evaluate the feasibility of using PEEK and PEEK composites as alternatives to metallic implants for rib reconstruction and to develop AI models capable of predicting stresses, strains, and deformations. Customized 3D models of a defective chest were reconstructed with implants made from PEEK, carbon fiber-reinforced PEEK (CFP), glass fiber-reinforced PEEK (GFP), and hydroxyapatite PEEK (HAP) as alternatives to titanium. FEM simulations were performed under lateral impact and sternal forces to extract mechanical responses, generating a comprehensive dataset used to train machine learning and deep learning regression models, including Linear Regression, Ridge Regression, Support Vector Regression, Decision Trees, Neural Networks, and LightGBM. Model performance was evaluated using R², MAE, MSE, RMSE, and computational efficiency. Results indicated that CFP 60% implants produced the lowest stress and strain levels on ribs and lungs, whereas pure PEEK and HAP 30% implants exhibited higher levels. GFP 30% and HAP 60% implants distributed tensile and compressive stresses similarly, though HAP 60% implants were prone to fracture due to excessive tensile stresses. AI models trained on FEM data achieved over 99.9% accuracy, demonstrating both predictive reliability and computational efficiency. These findings suggest that CFP (30% & 60%) and GFP (30% & 60%) composites are promising alternatives to titanium for rib reconstruction, and that integrating FEM with AI-based regression models can significantly optimize implant evaluation and design.