基于LVQ神经网络的微钙化分类方法

提出了一个基于自适应的学习矢量量化神经网络（LVQ）的乳腺肿瘤良恶性分类方法,在提取特征向量的基础上,对CC和MLO两种视图的良性和恶性数字化乳腺X光片图像进行训练和测试,并使用最佳分类率和平均分类率来分析分类结果。实验结果表明该方法对CC视图的图像的平均测试分类率为92.6%,而对MLO视图是93.18%。在微钙化分类系统中采用逻辑＂或＂的方式合并两种不同视图下的网络,可以获得的最佳分类性能是94.8%。

Classification Method of Micro-Calcifications Based on LVQ Neural Networks

A classification method of benign and malignant breast tumors based on adaptive LVQ（Learning Vector Quantization） neural networks is proposed,on the basis of extracting feature vectors,it trains and tests benign and malignant digitized mammograms of both CC and MLO views,and analyzes the classification results by using optimal and average classification rates.The experiment results show that the average test classification rate of the method is 92.6% for CC view images,93.18% for MLO view.In micro-calcification classification system,if logic OR is used to merge the networks under the two different views,the best classification performance that can be achieved is 94.8%.