摘要
由于癌细胞多形性,细针吸取细胞学检查存在局限性,辩识能力不足,造成误诊。为进一步提高乳腺癌辅助诊断的准确性,提出了一种基于概率神经网络的乳腺癌辅助诊断方法。首先建立基于概率神经网络的分类模型,其次确定网络的训练集和测试集,接着找出最优的径向基函数分布密度,最后计算5-折交叉验证的测试准确度,并对仿真方法和结果进行了检验。将仿真结果和检验结果与已有文献中所得出的结果进行对比分析,表明用概率神经网络进行乳腺癌的辅助诊断,具有准确度高,诊断用时少,易于实现等优点,说明了其在乳腺癌计算机辅助诊断方面的可行性和优越性。
In order to overcome the limitation of Fine Needle Aspiration Cytology, and to further enhance the ac-curacy of breast cancer auxiliary diagnosis, this paper proposed a computer-aided diagnosis method based on Probabi-listic Neural Networks. First, a classifying model based on probabilistie neural networks was established. Second, the samples were distributed to training set and test set. Third, after finding out the optimal radial basis density function of probabilistic neural networks, the accuracy of 5 -fold cross validation wass calculated. Finally, the method was checked and analysze with the results of simulation, and compared with other mothods. The results show that the probabilistic neural network has higher speed and better accuracy.
出处
《计算机仿真》
CSCD
北大核心
2012年第9期166-169,共4页
Computer Simulation
基金
国家自然科学基金资助项目(61179025)
湖北省教育厅重点项目(D20101202)
关键词
概率神经网络
乳腺癌
计算机辅助诊断
Probabilistic neural networks
Breast cancer
Computer-aided diagnosis
