基于神经网络的皮瓣供区闭合结果预测方法研究  

Prediction Method of Flap Donor Closure Based on Neural Network

作　　者：孙榕 纪小刚 李华彬 辛嘉铭 SUN Rong;JI Xiaogang;LI Huabin;XIN Jiaming(School of Mechanical Engineering,Jiangnan University,Wuxi 214122,Jiangsu,China;Jiangsu Key Laboratory of Advanced Food Manufacturing Equipment and Technology,Wuxi 214122,Jiangsu,China)

机构地区：[1]江南大学机械工程学院,江苏无锡214122 [2]江苏省食品先进制造装备技术重点实验室,江苏无锡214122

出　　处：《力学季刊》2024年第1期30-41,共12页Chinese Quarterly of Mechanics

基　　金：国家自然科学基金(52175234)。

摘　　要：针对研究皮瓣切口闭合机制的有限元法存在耗时长、专业性强等不足,本文提出了一种基于神经网络和有限元模拟的供区闭合应力与术后皮肤突起高度的快速预测方法.首先,构建考虑腿部供区组织纵剖面结构的超弹性有限元模型,对不同几何尺寸、不同组织厚度的切口闭合进行力学仿真分析,并建立神经网络数据集;搭建切口闭合模拟试验平台,采用数字图像相关法(Digital Image Correlation, DIC)验证有限元模型的可靠性;接着,以闭合模拟结果为样本的数据集作为输入,对BP (Back Propagation)、RBF (Radial Basis Function)、EBF(Elliptic Basis Function)三种模型进行训练优选,构建切口闭合预测模型;最后,采用预测模型扩充样本数据,结合Sobol灵敏度分析法探求各输入参数对切口闭合的影响.结果表明,EBF神经网络效果最佳,可有效预测切口闭合结果,切口短轴、长轴长度对闭合结果影响最大,皮肤厚度次之,脂肪厚度影响最低.同时,本文分析了各参数变化对闭合效果的影响,为腿部供区缝合手术提供参考依据.In view of the shortcomings of the finite element method to study the mechanism of flap incision closure,such as long time and strong expertise demand,this paper proposes a rapid prediction method of donor area closure stress and postoperative skin protrusion height based on neural network and finite element simulation.Firstly,a hyperelastic finite element model considering the longitudinal profile structure of the donor tissue was constructed,and mechanical simulation analysis was carried out for the incision closure of different geometric sizes and different tissue thicknesses,and a neural network dataset was established.A simulation test platform for notch closure was established,and digital image correlation(DIC)method was used to verify the reliability of the finite element model.Then,with the dataset of closed simulation results as the input,three models of BP(Back Propagation),RBF(Radial Basis Function)and EBF(Elliptic Basis Function)were trained and optimized,and the prediction model of cut closure was constructed.Finally,the prediction model was used to expand the sample data and the Sobol sensitivity analysis method was used to explore the influence of input parameters on the incision closure.The results showed that the EBF neural network has the best effect and could effectively predict the closure result of incision.The length of the short and long axis of incision have the greatest effect on the closure result,followed by the thickness of skin and the thickness of fat.At the same time,this paper analyzes the effect of different parameters on the closure effect,and provides a reference for the donor area suture surgery.

关 键 词：皮肤缝合 神经网络 预测模型 灵敏度分析 

分 类 号：O34[理学—固体力学] R6[理学—力学]