Inverse Design of Scaffolds for Bone Tissue Engineering using Artificial Neural Networks and Generative Additive...

Area of Science:

• Biomaterials Science
• Computational Biology
• Regenerative Medicine

Background:

• Scaffold design for bone tissue engineering is crucial for healing critical bone defects, often relying on complex simulations.
• Artificial neural networks (ANNs) show promise in simplifying scaffold design by predicting geometrical parameters.

Purpose of the Study:

• To develop an artificial neural network (ANN) for predicting gyroid scaffold design parameters using morphological characteristics.
• To investigate the prediction of anisotropic scaffolds for improved bone regeneration.
• To compare the performance of the ANN against a generalized additive model (GAM).

Main Methods:

• Generated a synthetic dataset of 6940 gyroid structures.
• Utilized 90% of the data for training and 10% for evaluation.
• Implemented a feature selection procedure to identify optimal predictive features.

Main Results:

• The ANN successfully predicted gyroid design parameters using only morphological features.
• ANN performance, measured by Pearson's correlation coefficients (0.53-0.82), significantly surpassed the GAM (0.41-0.51).
• The ANN demonstrated a lower mean absolute error compared to the GAM.

Conclusions:

• ANNs are effective tools for predicting scaffold design parameters in bone tissue engineering.
• The developed ANN approach, particularly for anisotropic scaffolds, shows potential for clinical applications.
• Further research is required to validate the clinical feasibility of this ANN-based scaffold design technology.