基于改进ReliefF与k-means算法的良恶性肺结节分类模型

肺结节是肺癌的症状.在CT图像中,肺结节的形状和大小常被用来进行肺癌的诊断,然而良性和恶性结节的鉴别对于疾病的治疗具有重要意义.由于良恶性结节的边缘纹理特征区别大,因此本文首先利用基于改进的边缘检测算子的灰度-梯度共生矩阵(GGCM)提取小梯度优势、灰度分布不均匀性、能量、灰度熵、梯度熵、混合熵、逆差距、相关性等肺部CT图像的14种纹理特征.然后利用改进的ReliefF算法去除作用小的特征,保留重要特征的特征权重值.最后将重要特征的权重值应用于改进距离度量准则的k-means算法中进行良恶性结节的分类.应用本文算法在LIDC数据集上实验,实验分析结果表明,14种纹理特征对于结节良恶性的分类能力并不相同,而灰度差、梯度差、能量、小梯度优势、相关性、灰度熵、混合熵、逆差矩的组合得到的良恶性肺结节分类效果最好,最终实现了良性结节83.46%,恶性结节95.02%的识别率,可在临床应用中辅助医生进行肺结节的良恶性诊断.

Classification Model of Benign and Malignant Pulmonary Nodules Based on Unbiased ReliefF and Weighted k-means Algorithm

Nodules are a symptom of lung cancer.In CT images,the shape and size of pulmonary nodules are often used in the diagnosis of lung cancer,but the differentiation of benign and malignant nodules is of great significance for the treatment of the disease.Due to the edge of the benign and malignant nodules texture feature difference is big, so at first, 14 texture features of lung CT images,such as small gradient advantage gray distribution heterogeneity,energy gray entropy gradient entropy,mixed entropy,deficit distance and correlation,are extracted by using the gray-gradient co-existence matrix(GGCM) based on the improved edge detection operator.Then,the improved ReliefF algorithm is used to remove the feature with little effect and retain the feature weight value of the important feature.Finally,the weight value of important features is applied to the K-means algorithm of improved distance measurement criterion for the classification of benign and malignant nodules.Using the classification model doing experiment on the LIDC data set,the experimental analysis results show that the 14 kinds of texture feature for the classification of benign and malignant nodules ability is not the same,the combination of gray difference,gradient difference,energy,small gradient advantage,correlation,gray entropy,mixed entropy and deficit moment obtained the best classification effect for benign and malignant pulmonary nodules.Finally,the recognition rate of benign nodules and malignant nodules was 83.46% and 95.02% respectively.And it can assist doctors in the diagnosis of benign and malignant pulmonary nodules in clinical applications.