摘要:目前临床上采用B超成像技术诊断脂肪肝大多是凭医生的经验判断,这种感性的认识常导致脂肪肝的误诊。本设计运用模式识别方法构建分类器,将提取的正常肝和脂肪肝B超图像纹理特征矢量,作为目标识别分类器的训练样本输入分类器,训练后的分类器最终可实现对肝脏B超图像正确识别。图像识别分类器设计是本课题研究的目的和重点,设计中采用误差反向传播(BP)神经网络、自组织特征映射神经网络(SOFM)及模糊C-均值聚类三种方法设计分类器,利用Matlab仿真比较各分类器达到的识别效果,最终确定对识别率最高的BP神经网络进行Visual
C++编程,并实现界面可视化。
关键词:脂肪肝;特征识别;神经网络;模糊C-均值聚类
Abstract: Nowadays, it is strongly dependent on the
doctor's empirical judgment by using the B-scan Ultrasonic images
to diagnose fatty liver. The perceptual recognition often leads to
the misdiagnosis of the fatty liver. Pattern recognition methods
have been used to design classifier which was trained to make
correct recognition and classification of liver ultrasound images.
The texture features vector which extracted form the normal liver
and fatty liver ultrasound images have been applied to the
classifier for training. In the design, the purpose and emphasis
focus on the designing of the image recognition classifier .In the
paper, the back propagation (BP) neural network, self-organizing
feature map (SOFM) neural network and fuzzy C-means clustering are
utilized to design the classifier in the platform of Matlab. Then
the best method is chosen from the three methods according to the
recognition rate. The back propagation neural network which is the
best one is designed in the platform of Visual C++. At last, the
user interface is implemented.
Key words: Fatty liver,
Feature recognition, neural
network, Fuzzy C-means clustering
目 录
联系QQ824797859