基于Adaboost的脑肿瘤患者存活周期分析

Life Cycle Analysis for Brain Tumor Patients Based on Adaboost

  • 摘要: 随着现代社会中脑部肿瘤发病率的日渐上升,脑肿瘤患者存活周期分析在临床中的意义也日渐凸显。为解决当前方法分析准确率不高的问题,该文提出一种基于Adaboost的脑肿瘤患者存活周期分析系统,首先对脑肿瘤患者的MR进行预处理、归一化、获取ROI和分割等处理,随后提取脑肿瘤患者的多序列MR的纹理特征以及进行以互信息为评价标准的特征选择并得到特征子集,最后搭建以Adaboost.R2为核心的分析模型,并利用特征子集完成分析模型的训练和调优,以完成肿瘤患者存活周期的分析。Brats2018训练数据上的交叉验证实验结果证实该系统的分析准确率优于Brats2018 challenge前3名的方法和传统回归分析方法。

     

    Abstract: With the increasing incidence of brain tumors in modern society, the analysis of the survival cycle of patients with brain tumors has become increasingly significant in clinical practice. In order to solve the problem of low accuracy of the current method, this paper proposes a life cycle analysis system for brain tumor patients based on Adaboost. Firstly, preprocessed magnetic resonance images and obtained its region of interest (ROI) and segmentation part, then extracted the texture features of multi-sequence MR for brain tumor patients, and performed feature selection using mutual information as the evaluation standard to obtain feature subset. Finally, this article builds an analysis model with Adaboost.R2 as the core method, and uses the feature subset to complete the training and tuning of the analysis model to complete the analysis of the survival period of tumor patients. The cross-validation experimental results on Brats2018 training data confirm that the analysis accuracy of this system is better than the Top3 methods of Brats2018 challenge and the traditional regression analysis methods.

     

/

返回文章
返回