14–17 Sept 2025
Palace of Culture and Science
Europe/Warsaw timezone

A machine learning-aided design framework for bone tissue engineering scaffolds

16 Sept 2025, 18:30
10m
Ratuszowa

Ratuszowa

Speaker

Pasquale Posabella (Warsaw University of Technology)

Description

Spherical-based architected scaffolds have gained increased interest in tissue engineering (TE) due to their curved architecture, which inherently reduces stress concentrations and enables controlled mechanical performance. However, designing and optimising such architectures remains challenging because of the intricate relationship between their structural and functional properties.
This work proposes an automated framework for optimising spherical-based porous scaffolds for bone TE. First, neural networks (NNs) were trained using data from finite element simulations to predict their stiffness and porosity based solely on geometrical features. Then, a genetic algorithm (GA) was coupled with the developed NNs to perform an inverse design, exploiting the predictive capability of the NN to tailor the microarchitecture based on the demands of mechanical and porosity. Next, validation through additional FE simulations confirmed the potential and highlighted the limitations of the presented framework. Finally, experimental validation was performed on 3D printed scaffolds.
The proposed tool serves a dual purpose. First, the NNs act as a surrogate model to instantly predict mechanical and porosity characteristics based solely on the scaffolds' microarchitecture. Secondly, it allows the inverse design of spherical-based scaffolds with combined normalised stiffness (between 0.060-0.226) and porosity (between 0.55-0.80). The precise control over porosity and mechanical properties, combined with the intrinsic anisotropic architectures, makes this tool suitable for generating optimised bone scaffolds, replicating the mechanical gradients between cortical and trabecular bone. Moreover, the frameworks can be used as standalone or with additional computational models, such as computational fluid dynamics, to assess nutrient transport in cell-culture experiments.
Further work will expand the tool’s ability to predict scaffold properties under large deformations (e.g., absorbed energy), providing a path for fully optimised bone TE scaffolds.

74734135559

Presentation materials

There are no materials yet.