Bayesian neural networks for probabilistic modeling of thermal dynamics in multiscale tissue engineering scaffolds

Abstract

Multiscale tissue engineering integrated with thermal dynamics exhibits a critical role as scaffolds in maintaining the structural integrity and functionality in fabrication applications. Thermal dynamics influence the properties such as thermal conductivity, the capacity of heat, and thermal expansion of materials to maintain the scaffold stability in various physiological conditions. However, the scaffold's heat distribution in a uniform manner is varied due to variations in pore size and geometrics. Additionally, variation in scales affects the thermal gradient impacts on the cell growth and integrity.

This paper proposes the 3D Scaffolds Probabilistic Weighted Bayesian Neural Network (3D-SP-WBNN) for tissue engineering with a multiscale scaffold model. It uses the 3D scaffolds for the multiscale design estimation. The weighted Probabilistic Bayesian Neural Network model is employed for estimating features in tissues and evaluating the cell growth and proliferation.

Thermal gradients measured were in between 1 °C and 4 °C for human bone, skin, and cartilage tissues. For low and moderate temperatures of 1 °C and 2 °C, the cell proliferation rates in cartilage tissues were 15–20 % per day. The scaffold design uses the Hybrid Hydrogel/PCL composite to achieve a higher proliferation rate of 25–35 % per day. The estimated forecasting achieves an accuracy range of 82–96 % for different cell densities and thermal conditions.