基于电子病历的乳腺癌群组与治疗方案可视分析  

Visual analysis of cohorts and treatments of breast cancer based on electronic health records

作　　者：徐敏[1] 王科[2] 戴浩然 罗晓博 余炜伦 陶煜波[3] 林海[3] XU Min;WANG Ke;DAI Haoran;LUO Xiaobo;YU Weilun;TAO Yubo;LIN Hai(Department of Medical Information,The First Affiliated Hospital of Zhejiang University School of Medicine,Hangzhou 310003,China;Department of Breast Surgery,The Second Affiliated Hospital of Zhejiang University School of Medicine,Hangzhou 310003,China;State Key Lab of CAD&CG,Zhejiang University,Hangzhou 310058,China)

机构地区：[1]浙江大学医学院附属第一医院医工信息部,浙江杭州310003 [2]浙江大学医学院附属第二医院乳腺外科,浙江杭州310003 [3]浙江大学CAD&CG国家重点实验室,浙江杭州310058

出　　处：《浙江大学学报（理学版）》2021年第4期391-401,共11页Journal of Zhejiang University（Science Edition）

基　　金：浙江省自然科学基金资助项目(LY18H160001)。

摘　　要：乳腺癌是当前最常见的恶性肿瘤之一,其电子病历数据可用于挖掘隐含规律,对治疗与预后分析有重要意义。通过与乳腺科医生合作,选择合适的预测模型和可视化方法,搭建了一个基于电子病历的乳腺癌群组和治疗方案可视分析系统。首先,对具有高维属性的病人进行降维和聚类处理,形成病人群组,并采用南丁格尔图、词云和时间轴可视化方法,直观展示病人群组间特征的差异;然后,用支持向量机(support vector machine,SVM)模型预测治疗方案,用平行坐标、矩阵热力图和分类图分别展示属性相关性、训练后的特征权重和预测结果;最后,用真实案例验证了系统在群组分析、治疗方案及病人属性关联分析中的有效性,从而较好地帮助医生选择合适的治疗方案。Breast cancer is one of the most common malignant tumors.It is important to analyze electronic health records(EHRs)and discover the patterns for breast cancer during the treatment and prognosis.By cooperating with physicians on breast cancer,this paper integrates well-designed visualization methods and an accurate prediction model to construct a visual analytics system for cohort analysis and treatment planning based on EHRs.For cohort analysis,the dimension reduction and clustering algorithm are first applied to patients with high-dimensional attributes to group them into the cohort,and Nightingale graph,word cloud,and timeline visualization methods are then used to explore the patterns in different cohort and correlation between features.For treatment planning,an SVM model is trained to predict the treatment plan for patients,and parallel coordinates,matrix heat map,and classification map are designed to explore the relation between features,the weights of features and the prediction result,respectively.Finally,case studies and expert interviews are presented to evaluate the effectiveness of cohort analysis and treatment planning in discovering the relationship between treatment plans and patient features and helping physicians in decision making.

关 键 词：乳腺癌 电子病历 病人群组 预测模型 可视分析 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]