影像组学在肺肿瘤良恶性分类预测中的应用研究

针对肺癌临床诊断中缺乏定量评估方法等问题,本研究采用影像组学方法构建基于支持向量机(SVM)的肺肿瘤良恶性分类预测模型。首先介绍了影像组学的定义、处理流程。实验样本选自公开数据集LIDC上的816例肺癌患者的CT影像数据。先采用中心池化卷积神经网络分割法提取感兴趣区(ROI),然后分别采用影像组学特征提取包Pyradiomics和FSelector特征筛选模型进行特征提取和特征降维,最后通过SVM构建肺肿瘤良恶性分类预测模型。模型对大于5 mm肺小结节的良恶性分类的预测准确率为80.4%,曲线下面积(AUC)的值为0.792,表明SVM分类器模型可以准确地判别大于5 mm的肺小结节的良恶性。

Application of Radiomics in Classification and Prediction of Benign and Malignant Lung Tumors

Aiming at the lack of quantitative evaluation methods in clinical diagnosis of lung cancer, a classification and prediction model of lung cancer based on Support Vector Machine(SVM) was constructed by using radiomics method. Firstly, the definition and processing flow of radiomics were introduced. The experimental samples were selected from 816 lung cancer patients on LIDC. Firstly, ROI was extracted by central pooling convolution neural network segmentation method. Then, Pyradiomics and FSelector feature selection models were used to extract features and reduce dimension. Finally, SVM was used to construct the classification and prediction model of lung tumors. The predictive accuracy of the model is 80.4% for the classification of benign and malignant pulmonary nodules larger than 5 mm, and the value of the area under the curve(AUC) is 0.792. This indicates that the SVM classifier model can accurately distinguish benign and malignant pulmonary nodules larger than 5 mm.