摘要
乳腺微钙化簇是早期乳腺癌的重要征象,计算机辅助的微钙化簇检测是医学影像领域的难题。为了提高检测系统的准确率,往往需要大量病灶标记,除了搜集样本本身的难度外,还需花费专家的大量时间。目前的研究工作很少涉及这个问题的解决方法。首次将基于主动学习的支持向量机技术应用到该领域,针对钙化簇感兴趣区域的特点,提出了选择训练集合的样本应该满足的基本条件。标准数据库上的实验证明,提出的方法能够大量地减轻样本标记的工作,并使乳腺癌微钙化簇检测系统的分类性能基本不变。
Clustered microcalcification is an important signal for breast cancer in the early stages. However, computer aided detection of microcaleification is a challenge in the field of medical imaging. To improve the performance of the de- tection system.a large amount of lesion labeling is essential. Besides the difficulty on collecting samples itself, it also takes experts much time for manual labeling. Few state-obthe-art techniques take into account this problem. W first ap- plied the techniques of active learning with SVM into this area to try to solve this problem. The basic conditions for the selected training set samples were proposed. The experiments on benchmark dataset show that our approach can reduce much works on labeling samples with holding the classification performance of the system of detecting interesting ROI regions.
出处
《计算机科学》
CSCD
北大核心
2010年第2期237-241,245,共6页
Computer Science
基金
陕西省教育厅科学研究计划基金(07JK381)
中国博士后科学基金(20070421126)资助
关键词
乳腺癌
计算机辅助检测
主动学习
支持向量机
Breast cancer,Computer aided detection,Active learning,Support vector machine