1) Develop validated neural network models that reproduce the dynamic behavior of menisci-tibio-femoral articulations.
2) Develop a simulation tool that can examine the tissue interdependencies of the knee during dynamic activity.
This research is funded by the National Science Foundation, Grant Number 506297, under the IMAG program for Multiscale Modeling. It is a collaborative effort that capitalizes on a diversity of expertise in areas such as clinical, experimental and computational biomechanics, nano-micro scale material modeling, finite element modeling, and neural networks.
Grant Numer: 506297
Principle Investigator: Trent Guess
Co-Investigators: Ganesh Thiagarajan, Amil Misra, Reza Derakhshani (University of Missouri - Kanas City), Lorin Maletsky (University of Kansas), Terence McIff (University of Kansas Medical Center)
Abstract from grant proposal
Dynamic loading of the knee is believed to play a significant role in the development and progression of tissue wear disease and injury. Macro level rigid body joint models provide insight into joint loading, motion, and motor control. The computational efficiency of these models facilitates dynamic simulation of neuromusculoskeletal systems, but a major limitation is their simplistic (or non-existent) representation of the non-linear, rate dependent behavior of soft tissue structures. This limitation prevents holistic computational approaches to investigating the complex interactions of knee structures and tissues, a limitation that hinders our understanding of the underlying mechanisms of knee injury and disease.
The objective of this project is to develop validated neural network models that reproduce the dynamic behavior of menisci-tibio-femoral articulations and to demonstrate the utility of these models in a musculoskeletal model of the leg. The specific aims of this study are:
Aim 1: Develop finite element (FE) models from micro-structure based constitutive methods that bridge the nano-micro scale behavior at the tissue level
Aim 2: Develop neural network (NN) based models that learn from FE simulation of dynamic behavior of menisci-tibio-femoral articulations
Aim 3: Validate the NN models within a rigid body dynamic model of a natural knee placed within a dynamic knee simulator
Aim 4: Demonstrate the utility of the NN models by placing them within a dynamic musculoskeletal model of the leg to study the interdependencies of the menisci and other knee tissues
Aim 5: Distribute the validated NN models of menisci-tibio-femoral dynamic response and contact pressure for use in any rigid body model of the knee or leg
The final product will be Neural Network (NN) models that conform to a modular application programming interface (API) that can be exported to any commercial integrated development environment (IDE) or in-house multi-body model. The NN models will be built upon a multi-scale approach and describe the non linear, rate dependent, non-homogenous dynamic response of menisci-tibio-femoral articulations in a computationally efficient modular package. The multi-scale modeling approach will be validated using a dynamic knee loading machine and the utility of the approach demonstrated by studying the interdependencies of menisci properties, tibio-femoral contact, and anterior cruciate ligament strain during a dual limb squat. A synergistic interdisciplinary team has been assembled to address the objective and aims of the proposed project comprising experts in rigid body dynamics and knee modeling, FE modeling, nano-micro scale material modeling, neural networks, and clinical and experimental biomechanics.
The proposed research will benefit society at large as the results of this work have potential applications to orthopedics, tissue engineering, and biomaterials. The work will also be a valuable asset to the musculoskeletal research community providing computational tools that may aid research in broad areas such as human movement, prosthetics, tissue engineering, sport injury, and disease.