基于随机森林的乳腺肿瘤细针穿刺辅助诊断  

Fine-needle aspiration diagnosis of breast neoplasms based on random forests

作　　者：孙伟[1] 张俊升[1] 邢培锐 

机构地区：[1]中国矿业大学信息与电气工程学院,江苏徐州221000

出　　处：《计算机应用》2015年第A02期143-145,169,共4页journal of Computer Applications

摘　　要：细针穿刺细胞学诊断是乳腺肿瘤早期诊断最常用的方法。为提高乳腺肿瘤细针穿刺诊断的准确率,提出了基于随机森林(RF)和支持向量机(SVM)的乳腺肿瘤细针穿刺辅助诊断方法。该方法利用乳腺肿瘤细针穿刺病例数据库,分别对随机森林(RF)、支持向量机(SVM)两种分类算法进行训练,并利用训练得到的分类模型对乳腺肿瘤进行诊断。仿真结果表明,采用RF分类器时,乳腺肿瘤诊断准确率达到95.96%,高于SVM分类器的94.71%,也高于学习向量化(LVQ)神经网络的91.51%及中人工神经网络的91.25%,且RF分类器准确率的稳定性优于SVM分类器,可靠性高。最终结果证明:采用RF分类器提高了乳腺肿瘤细针穿刺诊断的正确率和可靠性,为乳腺肿瘤细针穿刺临床诊断提供了更加先进有效的方法。Fine needle aspiration cytology diagnosis is the most commonly used method for early diagnosis of breast neoplasms. In order to improve the accuracy rate of fine-needle aspiration breast neoplasms diagnosis,this paper proposed the methods for fine-needle aspiration diagnosis of breast neoplasms based on Random Forests（ RF） and Support Vector Machine（ SVM）. Using breast tumor fine needle biopsy cases database,RF and SVM classification algorithms were trained to get the classification model for breast neoplasms diagnose. The simulation results show that by using RF classifier,the average classification accuracy rate of ten times six-fold cross validation reaches 95. 96%,which is better than the 94. 71% of SVM classifier,the 91. 51% of Learning Vector Quantization（ LVQ） neural network and the 91. 25% of artificial neural network.Furthermore,the stability of RF classifier accuracy is better than that of SVM classifier. The final results show that RF classifier improves the accuracy and reliability of fine-needle aspiration diagnosis of breast neoplasms and provides advanced and effective method for clinical fine-needle aspiration diagnosis of breast neoplasms.

关 键 词：乳腺肿瘤 随机森林 支持向量机 细胞穿刺 交叉验证 

分 类 号：TP391.77[自动化与计算机技术—计算机应用技术]