The Intelligent Detection Of Liver Cirrhosis And Its Imaging With New Parameters

Liver cirrhosis is one of the major diseases that endanger our society.The intelligent detection of liver cirrhosis and its imaging with new parameters have an important significance to clinical diagnosis.From normal liver to cirrhosis liver,its stifness has changed obviously.The elasticity of liver has become the focus of liver cirrhosis intelligent detection.The paper designs and develops an intelligent system for liver cirrhosis classification based on spectrum analysis.It extracts elastic features from radio frequency(RF)signal of living human liver using conventional ultrasonic probe.And it is expected to provide an effective diagnostic information for young doctors who are lack of experience.At the same time,the elastic features based on spectral analysis,as new parameters,are used to make elastography.In order to improve the accuracy of elastography,these elastic parameters are used to make model with Bayesian network.The posterior probability of the model is coded to Pseudo color,achieving an elastography to provide a new low-cost elastic imaging method.The main contents of the paper are:1.The paper designs and develops an intelligent detection system for liver cirrhosis to assist in the diagnosis.The RF signal is decomposed by three-layer wavelet packet using Daubechies16 based on time-frequency analysis.All sub-bands energy distributions are fitted by the quadratic polynomials based on the least squares method.The statistical features(mean value,standard deviation,skewness and kurtosis)of these fitted parameters can characterize the liver elasticity.Four kinds of classifiers(nearest neighbor classifier,naive bayesian network,support vector machine and random forest)are used to classify normal liver and cirrhosis.Compared with the texture features and RF time series features,the result based on elastic features shows the optimal accuracy(94.26 ? 5.23%)and specificity(94.7%).2.The paper proposes a new elastic imaging method using these elastic parameters without adding hardware cost of ultrasonic diagnostic apparatus.These parameters can obviously separate normal liver and cirrhosis.It can provide a elastic imaging for doctors with intuitive observation.3.In order to reduce the error rate,this paper proposes the Bayesian Network model based on the elastic features for the first time.The posterior probability of the model is coded to Pseudo color and it makes the pseudo-color image to characterize the liver elasticity.The method can not only separate normal liver and cirrhosis,but also distinguish between different degree cirrhosis to a certain extent.