基于灰色神经网络的骨折愈合应力预测

在骨折愈合应力准确预测问题的研究中,骨折愈合受许多因素的影响,应力是主要因素之一。为了加快骨折愈合的速度和提高质量,及时了解骨折愈合过程中应力的变化趋势并调整加力的大小,就显得很重要。然而应力的变化是不确定的,用传统的观察法医生很难确定次日应力的大小,用单一的GM(1,1)模型预测精度也不高。根据神经网络能有效修正灰色预测模型的思路,提出了基于灰色系统理论及径向基神经网络的组合预测模型。先用灰色系统理论中的GM(1,1)模型,用已有的实测应力数据对次日的骨折断面应力进行预测,然后用实测值与预测值的差值训练神经网络,从而可以对灰色预测的值进行修正。实验结果表明:提出的应力预测的模型获得较高的预测精度,说明组合预测模型效果优于单一的灰色预测模型。

Forecast of Fracture Healing Stress Based on Grey Neural Network

Fracture healing is affected by a lot of factors,and stress is one of the main reasons.In order to accelerate the speed of fracture healing and improve its quality,it is very important to be informed the change trends of stress and adjust the size of the external force during fracture healing process in time.However,the change of stress is uncertain,it is difficult to determine the size of the stress for doctor with the traditional observation,and the prediction accuracy of single GM(1,1) model is also not high.Based on the thought that neural network can effectively modify grey prediction model,a combined forecasting model was proposed,based on the grey system theory and RBF neural network.First,GM(1,1) model,from the grey system theory,was used.We predicted the fracture section stress of next day by the existing measured data of actual stress.Then the difference value between actual and forecast value was used to train the neural network.Thus we can amend grey prediction value.Experimental results show that the stress forecasting model is feasible and can acquire higher precision of prediction,what’s more,the effect of hybrid forecast model is better than a single grey prediction model.