Font Size: a A A

The Research For Detection And Recognization Degree Of Early Liver Fibrosis Based On ARMA Model

Posted on:2018-07-22Degree:MasterType:Thesis
Country:ChinaCandidate:T YiFull Text:PDF
GTID:2334330536978130Subject:Engineering
Abstract/Summary:PDF Full Text Request
Chronic liver injury is usually a common outcome of liver fibrosis,and early liver fibrosis which is irreversible,developing into liver cirrhosis or even liver cancer is irreversible.Therefore,dynamic monitoring and grading the stages of early is very significant.In this paper,ultrasound RF time series was used as the data source and the rat liver was used the experimental object.We investigated the possibility of applying ARMA model to ultrasound RF time series and validate the effectiveness of the ARMA model parameters for the characterization of ultrasound tissue on the in-vivo of rats that had diffierent stages of liver fibrosis.Firstly,ultrasound RF time series was detected and pretreated to satisfy the conditions of establishing ARMA model.Then,the parameters of ARMA model with three different parameter estimation algorithms were used to distinguish the stages of early liver fibrosis in different classification.Lastly,the effects of different parameter estimation algorithms and order of ARMA model and different ROI size on classification accuracies were analyzed.We used three different sizes(10 ×60,20×70,25×75 pixels)to conduct the same experiments for analysizing the effects of ROI sizes.When ROI size was set 10 ×60 pixels,ARMA(4,3)model based on RLS_LS parameter estimataion algorithm assessed and staged liver fibrosis using Random Forest classifier.The average accuracies and STD(Standard Deviation)were 93.79% ±2.68(S ?1),83.48% ±5.42(S ?2),90.00% ±5.91(S ?3)and 94.09% ±3.61(S ?4)respectively for ARMA(4,3)model.The average accuracy and STD of the parameters of ARMA(4,3)for differentiating between grade S0,grade S1-S3 and grade S4 fibrosis were 90.61% ± 4.38.Experimental results showed that the parameters of ARMA(4,3)model achieved the classification averaged accuracy and STD of 75.85% ± 4.41 for differentiating the five stages of liver fibrosis.Compared with the effect of different ROI size on the classification accuracy,it was found that there had no significant effect.Finally,based on MFC platform of VS2010,the application system for differentiating early liver fibrosis was development,which provides a new way of thinking for the detection and montition of early fibrosis in human liver.
Keywords/Search Tags:Early fibrosis, ARMA model, ultrasound RF time series, random forest
PDF Full Text Request
Related items