A machine learning model exploring creep performance of dental composites.

Area of Science:

• Materials Science
• Biomaterials Engineering
• Computational Biology

Background:

• Red blood cell (RBC) viscoelastic creep is crucial for dimensional stability and clinical performance.
• Analyzing factors influencing RBC creep is complex due to material variations and testing protocols.

Purpose of the Study:

• Develop a robust machine learning (ML) model to predict RBC creep behavior.
• Identify critical compositional factors influencing RBC creep performance.

Main Methods:

• Measured compressive creep deformation and recovery for 5 RBCs under dry/wet conditions.
• Encoded RBC compositional data into 17 binary indices for ML model training.
• Evaluated 40 ML models for predictive accuracy.

Main Results:

• Creep was composition-dependent and increased significantly with water storage.
• The ExtraTrees ML model achieved R²=0.9999 in predicting strain, permanent set, and creep recovery.
• ORMOCER and SiO2 were identified as the most influential monomer and filler components, respectively.

Conclusions:

• The ML model effectively predicts RBC creep performance and identifies key compositional drivers.
• This approach can guide the development of advanced dental composites with enhanced clinical durability.