Optimizing Peripheral Nerve Regeneration using Computational Intelligence based T

描述（申请人提供）：周围神经损伤是每年影响大量患者的常见病。组织工程已成为开发周围神经再生替代神经移植物的有力途径。由于周围神经再生的组织工程策略涉及各种可能的变量组合，因此有必要开发有效的工具来识别最佳组织工程策略并根据这些组织工程策略预测周围神经再生的实验结果。一些研究小组已经应用人工神经网络和决策树来获得组织工程策略预测的最佳模型配置。对于基于决策树的方法，很难判断哪一种分类树比另一种分类树更好。此外，使用决策树算法的预测系统缺乏随时间积累学习经验的能力。另一方面，人工神经网络（ann）具有一些显著的特性，但只有连接权是用固定的拓扑来训练的。对于每种特定的组织工程策略，很难预先找到最佳的固定拓扑结构。本文提出了基于群体智能的进化人工神经网络技术来解决这一挑战。本项目将采用蚁群优化（Ant Colony Optimization， ACO）和粒子群优化（Particle swarm Optimization， PSO）两种基于群体智能的方法来训练人工神经网络模型。更具体地说，将使用蚁群算法优化人工神经网络模型的拓扑结构，而使用粒子群算法根据优化后的拓扑结构调整人工神经网络模型的连接权值。这种基于SWarm智能的人工神经网络强化学习方法（SWIRL-ANN）可以自动同时进化人工神经网络的拓扑结构和连接权，从而为周围神经再生的组织工程策略提供最优分类器。该研究项目将包括以下几个阶段：目标1：使用基于SWarm智能的神经网络强化学习方法（SWIRL-ANN）分析和预测系统预测周围神经再生的组织工程策略。目的2：利用基于SWIRL-ANN的分析预测系统，验证新型未知组织工程神经移植物在大鼠坐骨神经损伤模型中周围神经间隙桥接的有效性。

DESCRIPTION (provided by applicant): Peripheral nerve injuries are common diseases that affect a large amount of patients every year. Tissue engineering has emerged as a powerful approach for developing alternative nerve grafts for peripheral nerve regeneration. Since tissue engineering strategies in peripheral nerve regeneration involve various possible combinations of variables, it is necessary to develop efficient tools to identify optimal tissue engineering strategies and predict the experimental results based on these tissue engineering strategies for peripheral nerve regeneration. Some research groups have applied artificial neural networks and decision trees to obtain the best model configuration for the prediction of the tissue engineering strategies. For the decision trees based methods, it is hard to tell which classification tree is better than the other. Furthermore, the prediction system using the decision tree algorithm lacks the capability of accumulating the learning experience over time. On the other hand, Artificial Neural Networks (ANNs) exhibit some remarkable properties, but only the connection weights are trained with fixed topology. It is hard to find the best fixed topology in advance for each specific tissue engineering strategy. In this proposal, swarm intelligence (SI) based evolving ANNs technique is proposed to tackle this challenge. Two swarm intelligence based methods, Ant Colony Optimization (ACO) and Particle Swarm Optimization (PSO), will be applied in this project to train the ANN model. More specifically, ACO will be used to optimize the topology structure of the ANN models, while the PSO is used to adjust the connection weights of the ANN models based on the optimized topology structure. For this SWarm Intelligence based Reinforcement Learning method for ANNs (SWIRL-ANN) system, both topology and connection weight of artificial neural networks can be evolved automatically and simultaneously so that an optimal classifier for tissue engineering strategies in peripheral nerve regeneration can be achieved. The research project will include the following phases: Aim 1: Predict tissue engineering strategies in peripheral nerve regeneration using SWarm Intelligence based Reinforcement Learning method for ANNs (SWIRL-ANN) analytical and prediction system. Aim 2: Validate the efficacy of novel unknown tissue engineered nerve grafts as predicted by using SWIRL-ANN based analytical and prediction system for bridging peripheral nerve gaps in rat sciatic nerve injury model in vivo. PUBLIC HEALTH RELEVANCE: Tissue engineering has emerged as a powerful approach for developing nerve grafts for peripheral nerve regeneration. Since tissue engineering strategies in peripheral nerve regeneration involve various possible combinations of variables, it is necessary to develop efficient tools to identify optimal tissue engineering strategies and predict the experimental results based on these tissue engineering strategies for peripheral nerve regeneration. In this proposal, swarm intelligence (SI) based evolving artificial neural networks (ANNs) technique is proposed to tackle this challenge. The proposed research will be helpful to efficiently develop tissue engineered products for tissue and organ replacement.