As technology advances towards complex tissue engineering methods, scientists are hitting a ceiling on how accurate computer models can simulate tissue grafts grown to repair a patient's damaged tissues. The National Institute of Health (NIH) has recently awarded a grant to researchers at Stevens Institute of Technology to pursue novel approaches to engineering peripheral nerve tissues that incorporate lessons from swarm intelligence into computer simulations. Dr. Xiaojun Yu, Associate Professor of Biomedical Engineering, and Dr. Yan Meng, Assistant Professor of Computer Engineering, are collaborating to break through the existing technology barrier and develop smarter nerve tissue grafts.
"Since tissue engineering involves many variables, it is necessary to develop efficient tools to identify optimal strategies and predict experimental results based on these strategies for peripheral nerve regeneration," Dr. Yu reports. Computerized abstractions of the nervous system, called artificial neural networks (ANN), create models that strongly resemble real neuron behaviors and are commonly used to simulate tissue growth in vitro. However, conventional ANN also has limits because it cannot convey sufficient complexity to accurately predict tissue engineering results in cases like peripheral nerves.
To evolve ANN techniques in tissue engineering, the Stevens scientists propose to apply knowledge from the field of swarm intelligence (SI) to ANN models to improve their accuracy. A "swarm" is any community of seemingly non-intelligent agents, lacking centralized control, that nevertheless exhibit collective behaviors that appear "intelligent." The most common example of a swarm is the ant colony, which collectively makes astonishingly efficient decisions about finding food and defending the nest. SI is an interdisciplinary field that promotes study of these types of systems that demonstrate swarming behavior. The goal of SI research is to understand what causes high-functioning global patterns to emerge out of apparently unsophisticated individual actions. The SI-based methods are population-based stochastic searching methods mainly developed for system optimization, such as Ant Colony Optimization (ACO) and Particle Swarm Optimization (PSO).
The primary application of this NIH grant is to target tissue engineering of peripheral nerves, which extend from the central nervous system to link the brain and spinal cord to other parts of the body, like the muscles and skin. As these nerves are fragile and not protected by bone, they become easily damaged as a consequence of trauma or stresses on the body. Peripheral nerve injuries are common occurrences in trauma patients, resulting in local loss of sensory or motor functions and often requiring surgery to promote healing and regeneration of nerve endings.
With this NIH grant, the Stevens team is exploiting the elements of the complex, yet intricately designed peripheral nerve structures. In this project, they propose to generate and tune the structure and weights of ANNs using two SI-based optimization methods ( ACO and PSO). Their results are leading to the development of a Swarm Intelligence based Reinforcement Learning method for ANNs (SWIRL-ANN) system.
"With the development of advanced experimental tools and methods, data analysis has become a bottleneck in bioscience and biomedical innovation," says Dr. Meng, and the researchers are cracking the technology barrier with this new computational modeling tool. SWIRL-ANN works by first enabling the team to accurately predict the results of complex tissue engineering strategies, without needing to spend time and resources to conduct physical experiments by trial and error. Once optimal strategies are identified, the investigators will then validate the efficacy of novel unknown tissue engineered nerve grafts through in vivo sciatic nerve injury repairs.
The results of this research will serve as the foundation for future work aimed at optimizing tissue engineering approaches, and will provide important insight into classification and prediction of peripheral nerve regeneration. Beyond this preliminary application, the SWIRL-ANN model offers a generic, powerful tool for advanced research in the biological sciences.By combining the knowledge in artificial intelligence, tissue engineering, biomaterials, and neural defect surgery, Dr. Yu and Dr. Meng's results will contribute broadly to fundamental understanding in the field.