CT图像中肿大淋巴结肺癌转移分类方法

Classification of Tumid Lymph Nodes Metastases and Non-Metastases from Lung Cancer in CT Image

为解决肺癌N分期中胸部CT难于对肿大淋巴结是否癌转移进行评价的问题,寻求能够有效表示淋巴结病理特性的图像特征,实现对肿大淋巴结癌转移快速准确地判别。该文采取交互式分割从CT图像中提取出肿大淋巴结;直接计算淋巴结的多分辨率直方图得到200维空间信息特征样本集;利用具有处理高维数据集优势的支持向量机(SVM)构造分类器;用测试集对经训练的SVM分类器进行测试以评价分类性能。经96例病例实验结果表明:100个淋巴结图像的200维特征计算用时1.91s,SVM分类器训练测试用时1.36s,敏感性76%,特异性64%,准确度70%,接受者操作特性曲线(ROC)下面积(AUC)0.6525。高维图像空间信息特征能够有效表示淋巴结特性;没有考虑医学征象进行肿大淋巴结癌转移定性诊断的准确度就达到了70%,同时分类速度比传统纹理算法提高了约10倍。

In order to solve the low accuracy diagnosis of metastases and non-metastases tumid lymph nodes in the lung cancer Nstage with chest CT images, effective image features of lymph nodes need to be found for quickly and accurately differentiating metastases and non-metastases tumid lymph nodes. First, tumid lymph nodes are extracted from chest CT images using interactive segmentation. Second, the multi-resolution histograms of tumid lymph nodes are directly calculated to receive a high-dimensional features sample set with spatial information. Then the classifier for differentiating metastases and non-metastases tumid lymph nodes is constructed with making full use the advantage of SVM which is good at dealing with high dimensional data sets. Finally, the performance of classification is evaluated by testing the trained SVM with the test sample set. The test results by 96 cases show that it takes 1.91 s for computing 200 dimensional features of 100 lymph nodes, 1.36 s for training and testing the SVM classifier. Receiver Operating Characteristic （ROC） analysis of the classification performance shows that the sensitivity is 76%, specificity is 64%, accuracy is 70%, and the Area Under Curve （AUC） is nearly 0.6525. Image spatial information can effectively express the characteristics of lymph nodes, the classification accuracy of metastases and non-metastases tumid lymph nodes is up to 70% without medical signs, and the classification speed is about 10 times than traditional texture methods. It provides a feasible, simple, objective method for improving the accuracy of the lung cancer N stage in medical imaging diagnosis.