基于小波变换的乳腺肿瘤B超图像识别的研究

通过小波变换的原理对正常乳腺B超图像和病变乳腺B超图像进行小波分解，对图像进行小波去噪处理，再对图像进行小波特征提取。通过人工神经网络的方法对图像的特征参数进行统计分析，得出正常的乳腺B超图像和发生病变的B超图像之间的区别，从而判断哪些图像发生病变。仿真实验表明，该方法相对于医生凭经验判断有更高的准确率。结论：采用小波变换方法将图像分解、去噪并提取出来的特征参数可以有效地将两类图像区分开来，医生根据量化特征参数进行诊断，提高乳腺肿瘤临床诊断的准确率。

Research on B-ultrasonic Image Recognition of Breast Tumor Based on Wavelet Transformation

In this paper, we decompose the B-ultrasonic images of normal breast and distempered breast by using the theory of wavelet transformation, then carry on wavelet de-noising processing and wavelet feature extraction, and at last sta- tistically analyze the feature parameter of the image though the methods of Artificial neural network so as to get the differ- ences of the two kinds of B-ultrasonic images, from which we can know which images is abnormal. Experimental results show that this method is more accurate than the doctors＇ experience judgement. Conclusion： By using the method of Wavelet Transformation, feature parameter extracted from Image Segmentation and image de-noising can distinguish the two kinds of images efficiently. Doctors＇ diagnosis which is made according to the quantized feature parameter can improve accuracy rate of clinical diagnosis of Breast Tumor.