Abstract: A novel approach based on Dynamic Multiple Classifiers Selection（DMCS） ensemble algorithm is proposed based on the characteristics of diversity and inhomogeneity of the nodule lesions. The feature space is randomly divided into a number of feature subsets. Specifically, for different kinds of subset, more appropriate base classifier is selected based on their distributions. Then the results of the classifiers are ensembled for final decision. The experimental results demonstrate that the proposed algorithm achieves better performance than the typical and traditional nodule detection approaches.

Key words: lung nodule, dynamic multiple classifiers selection, integration algorithm, computer aided detection

摘要： 针对肺结节病灶数据具有多样性及异质性特点，提出了动态多分类器选择集成算法（Dynamic Multiple Classifiers Selection，DMCS），将特征空间随机划分为若干特征子集，针对每个特征子集样本分布不同，对不同的特征子集选择适合的基分类器，最后进行集成学习。实验表明，该算法比目前有代表性的肺结节检测病灶分类算法具有更好的稳定性和检测性能。

关键词: 肺结节, 动态多分类器选择, 集成算法, 计算机辅助检测