模糊遗传人工神经网络算法提取乳腺微钙化点的效果  

Effect of Fuzzy Genetic Artificial Neural Networks Agorithm Extracting Microcalcification

作　　者：鲁雯[1] 张光玉[2] 

机构地区：[1]同济大学电子与信息工程学院,上海200092 [2]泰山医学院放射学院,山东泰安271016

出　　处：《中山大学学报（医学科学版）》2008年第2期216-220,234,共6页Journal of Sun Yat-Sen University:Medical Sciences

基　　金：国家自然科学基金(60572113)

摘　　要：【目的】微钙化点是早期乳腺癌的重要征象之一,本研究联合运用遗传算法、模糊数学和人工神经网络,建议一种乳腺微钙化点提取的新方法,为乳腺病变的自动识别提供前期处理,为早期乳腺癌的临床诊断提供帮助。【方法】首先利用随机方法产生大量的样本,然后,利用模糊遗传算法对产生的随机样本进行分类,将分类后的样本输入人工神经网络进行训练,将310幅乳腺图像的感兴趣区域输入训练后的人工神经网络分类器进行分类。【结果】与微钙化点提取方面的同类文献相比较,结果表明该算法在相同误检率下得到较高的阳性检出率。【结论】研究表明综合运用遗传算法、模糊数学和人工神经网络进行乳腺微钙化点提取比单纯运用人工神经网络提取效果好。[ Objective ] Microcalcification is one of the most important characteristics of early breast tumors. In this paper, we proposed a new method of microcalcification detection by integrating genetic algorithm, fuzzy mathematics and artificial neural networks. The method could provide preprocessing for automatic recognition of breast cancers, and assist doctors to diagnose early breast cancer. [ Methods ] A lot of random training samples were firstly produced; then, these samples were classified into the background and microcalcifications using the fuzzy genetic method. Finally, the 310 regions of interest were classified into the background and microcalcifications using the trained neural networks. [Results] Compared with similar literature about microcalcification detection, we obtained better positive detection ratio with the same false detection ratio. [ Conclusions] Experimental results demonstrate that our method obtain better extraction effect by integrating genetic algorithm, fuzzy mathematics and neural network, compared with the method simply using artificial neural network.

关 键 词：遗传算法 人工神经网络 微钙化点 模糊数学 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]