基于小波变换的脂肪肝B超图像识别

目的探讨超声图像后处理的临床诊断价值.方法采用小波变换方法对脂肪肝和正常肝的B超图像进行多分辨分析,对小波变换系数进行统计分析,提取变换系数的均值和方差参数,根据提取的特征参数采用概率神经网络对图像进行模式识别.结果对40幅正常肝和40幅脂肪肝图像中的感兴趣区域提取特征参数,训练后的网络对脂肪肝和正常肝的正确识别率分别为88%和85%.结论采用小波变换方法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,提高了脂肪肝临床诊断的准确率.

Fatty liver ultrasonic image recognition based on wavelet transform

Objective To investigate the value of medical image postprocessing in diagnosing fatty liver. Methods The wavelet transform was used in analyzing liver ultrasonic image based on multl-resolution analysis. Statistical features such as mean and standard deviation were extracted from transformation coefficients, then the above statistical features were applied for texture classification by probabilistic neural network. Results Experimental results showed that the method could achieve about 88% identification rate with fatty liver and about 85% with normal liver. Conclusion The feature data extracted from ultrasonic images is useful for increasing the accuracy of clinical diagnosing fatty liver.