基于不同常数项模糊支持向量机的脑变形矫正模型  

A Brain Shift Correction Model Based on the Fuzzy Support Vector Machines with Different Constant Term

作　　者：王伟[1] 章琛曦[1] 

机构地区：[1]复旦大学数字医学研究中心,上海200032

出　　处：《生物医学工程学杂志》2010年第6期1360-1364,共5页Journal of Biomedical Engineering

基　　金：国家自然科学基金资助项目(30600639);国家863计划资助项目(2009ZZ02Z415);上海市重点学科B112资助项目;上海市科委国际合作课题资助(09410702800)

摘　　要：术中脑组织变形是影响神经外科手术导航(IGNS)系统精度的主要原因。本文试图建立脑组织变形量和影响脑组织变形的因素之间的统计学习模型,并用这种模型来预测脑组织变形情况。脑组织变形预测可看成是一个函数回归估计问题:脑组织形变量为函数的输出值,而相应的影响脑组织变形的因素作为函数输入值。问题的最终目标是寻找一个有良好推广能力的从影响脑组织变形的因素到脑组织变形的映射。本文采用不同常数项模糊支持向量机(FSVM)对脑组织变形量和影响脑组织变形的因素建立脑组织变形矫正模型。采用该方法对10组临床数据集进行训练,得到脑组织位移与开颅方向、手术部位等多维数据的关系模型,最后用留一法检验该模型,结果显示,9组数据集的预测误差小于等于1 mm。研究结果表明:基于不同常数项FSVM的脑变形矫正模型可以用来进行术前脑组织变形预测,并获得临床可接受的预测精度,能应用到IGNS手术导航中。Brain shift contributes mostly to the error of prediction in ImageGuided Neurosurgery（IGNS）.In order to solve this problem,we build a statistical learning model between the quantity of the brain shift and we factors that impinge on the brain shift,and we predict the brain shift by using this model.The prediction of the brain shift can be regarded as an approach to estimation of regression function： the quantity of the brain shift is the output value of the function,while the corresponding factors that affect the brain shift can be taken as the input value of the function.In this study,we employ the fuzzy suppert vector machines（FSVM） with different constant term to build a brain shift correction model between the quantity of the brain shift and the factors that affect the brain shift.By taking 10 clinical data sets and employing the novel predicting method,we trained the relational model of the multi-dimensional data for the brain tissue displacement,the direction of surgical operation,and the operative site,etc.The results of validating the model by the leave-one-out method unveil that the approach recapitulated 90% of the shift,thus indicating that the correction model based on the FSVM with different constant term can be used to predict the brain shift with clinically acceptable accuracy.Therefore,the model can be applied to IGNS.

关 键 词：模糊支持向量机 脑变形 神经外科手术导航 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]