摘要
目的分析支持向量机(SVM)对结肠腺瘤高级别上皮内瘤变的发生预测效果。方法随机抽取2008—2010年北京大学人民医院消化科行结肠息肉高频电凝切除术患者157例,收集每例患者临床及内镜相关的12个指标,利用SVM及Logistic回归分析两种方法分别对结肠腺瘤高级别上皮内瘤变的发生进行预测。在应用SVM预测部分,首先随机抽取50例作为训练集,利用LIBSVM-2.88软件建立预测模型,再从剩余病例中随机抽取3组,作为3个测试集,应用训练集所得的模型对3个测试集分别进行预测,得出平均预测正确率及特异度、敏感度。在应用Logistic回归分析部分,全部157例患者均纳入Logistic回归模型进行分析。最后对SVM及Logistic回归模型两种方法的预测结果进行比较。结果 SVM预测分类结果 3个测试集平均预测正确率为(92.6±3.3)%,平均敏感度为(80.6±17.3)%,平均特异度为(94.8±0.6)%。Logistic回归模型预测正确率为90.4%。结论应用SVM建立的预测模型在小样本的基础上对结肠腺瘤发生高级别上皮内瘤变获得了较好的预测效果。
Objective To analysis effect of support vector machine （SVM） to predict the high grade intraepithelial neoplasia of colonic adenoma. Methods SVM is a kind of recognized technique which is based on statistical learning theory and the principle of structural risk minimization. SVM can obtain the optimum result from limited samples. Clinical doctors are not familiar with this technique. 157 patients from Peking university people＇s hospital who underwent colonoscopy were selected randomly and 1：2 clinical and endoscopic data were collected of every patient. In the part of SVM research, we have used the data of 50 randomized patients who selected as training set to build a SVM model, by this model, we have predicted the oc- currence of high grade intraepithelial neoplasia in three testing sets, calculated the average predicting accuracy, sensitivity, specificity. In the meantime,we also built a logistic regression model used the data of total 157 patients and compared the predicting effect with SVM. Results The average predicting accuracy, sensitivity, specificity of SVM model is （ 92. 6 ± 3.3 ） %, （ 80. 6 ±17.3 ） %, （ 94. 8 ± 0. 6） %. The predicting accuracy of logistic regression analysis is 90. 4%. Conclusion SVM achieved satisfactory effect in predicting the occurrence of high grade intraepithelial neoplasia in colonic adeuoma.
出处
《中国实用内科杂志》
CAS
CSCD
北大核心
2013年第11期872-875,I0001,共5页
Chinese Journal of Practical Internal Medicine
基金
首都医学发展科研基金(3031125)
