基于多分类AdaBoost改进算法的TEE标准切面分类

发布时间：2018-08-23 10:10

【摘要】：针对超声图像样本冗余、不同标准切面因疾病导致的高度相似性、感兴趣区域定位不准确问题,提出一种结合特征袋(BOF)特征、主动学习方法和多分类AdaBoost改进算法的经食管超声心动图(TEE)标准切面分类方法。首先采用BOF方法对超声图像进行描述;然后采用主动学习方法选择对分类器最有价值的样本作为训练集;最后,在AdaBoost算法对弱分类器的迭代训练中,根据临时强分类器的分类情况调整样本更新规则,实现对多分类AdaBoost算法的改进和TEE标准切面的分类。在TEE数据集和三个UCI数据集上的实验表明,相比AdaBoost.SAMME算法、多分类支持向量机(SVM)算法、BP神经网络和AdaBoost.M2算法,所提算法在各个数据集上的G-mean指标、整体分类准确率和大多数类别分类准确率都有不同程度的提升,且比较难分的类别分类准确率提升最为显著。实验结果表明,在包含类间相似样本的数据集上,分类器的性能有显著提升。
[Abstract]:Aiming at the problem of redundancy of ultrasonic images, high similarity of different standard sections caused by diseases and inaccurate localization of regions of interest, a new feature bag (BOF) feature is proposed. Active learning method and improved multi-classification AdaBoost algorithm for transesophageal echocardiographic (TEE) standard sectional classification. First, the ultrasonic image is described by BOF method, then the most valuable sample is selected as the training set by active learning method. Finally, in the iterative training of weak classifier using AdaBoost algorithm, According to the classification of temporary strong classifier, the updating rules of samples are adjusted, and the improvement of multi-classification AdaBoost algorithm and the classification of TEE standard section are realized. Experiments on TEE data sets and three UCI datasets show that compared with AdaBoost.SAMME algorithm, the (SVM) algorithm of multi-classification support vector machine has BP neural network and AdaBoost.M2 algorithm, and the G-mean index of the proposed algorithm on each data set is better than that of AdaBoost.SAMME algorithm. The overall classification accuracy and the classification accuracy of most categories are improved to some extent, and the more difficult classification accuracy is the most significant. The experimental results show that the performance of the classifier is significantly improved on the dataset containing similar samples between classes.
【作者单位】：中国科学院成都计算机应用研究所;中国科学院大学;
【基金】：四川省科技支撑计划项目(2016JZ0035) 中国科学院西部之光项目~~
【分类号】：R445.1;TP391.41

【参考文献】