INTRODUCTION:
Autografts are the gold standard for ligament replacement but have notable disadvantages. Tissue-engineered implants, particularly electrospun scaffolds, offer a promising alternative by replicating the extracellular matrix's morphology [1]. Machine learning (ML) can optimise the electrospinning process, creating ligament implants with morphology and mechanical properties similar to native tissues [2]. Studying the in vivo hyper-elastic behavior of ligaments through motion capture and musculoskeletal models can further inform biomimetic construct design [3]. This research aims to develop a manufacturing optimisation methodology using ML models and experimental biomechanics to create biomimetic ligament implants.
METHODS:
Polyvinyl alcohol (PVA) scaffolds were produced by systematically modifying the polymer concentration and production parameters. Both 2D and 3D scaffolds were characterised morphologically via scanning electron microscopy and mechanically through tensile testing. Data from 2560 observations informed 20 ML models to predict fibre diameter and inter-fibre separation. Additionally, 28 ML models predicted mechanical properties, including Young's modulus and ultimate tensile strength. A musculoskeletal knee model, combined with kinematic data from 12 young participants, estimated the in vivo biomechanics of the anterior cruciate ligament (ACL).
RESULTS AND DISCUSSION:
Decision Trees and Rule-Based Models generated a visual route to optimise the electrospinning process, achieving a morphology prediction accuracy of 0.868. Cubist models were most accurate for predicting mechanical properties, with an R² of 0.93. Crosslinked triple-twisted/braided filament scaffolds replicated the hyper-elastic behaviour of the native ACL effectively, showing R² values of 0.971 and 0.999 when using Mooney Rivlin and non-linear string-based models, respectively.
CONCLUSIONS:
PVA electrospun scaffolds, optimised using Decision Trees and Rule-Based Models, successfully replicated the morphology and hyper-elastic behaviour of natural ACLs.
REFERENCES:
- Roldán, E. et. al. Frontiers in Bioengineering and Biotechnology 2023, 11, doi:10.3389/fbioe.2023.1160760.
- Roldán, E. et. al. Frontiers in Physics 2023, 11, doi:10.3389/fphy.2023.1112218.
- Roldán, E. et. al. Materialia 2024, 33, 102042, doi:10.1016/j.mtla.2024.102042.
