Preliminary Application Of Artificial Neural Network In Ultrasonographic Qua Ntitative Diagnosis Of Hepatic Fibrosis

The liver fibrosis had majority had in each kind of cause of disease causes the chronic liver disease, in these, 20%~40% patients will develop to liver cirrhosis and even primary hepatocellular carcinoma. So it has been a crucial question to block or retroconversion liver fibrosis for development. Therefore it becomes a important study to early diagnosis of liver fibrosis.PurposeTo evaluate the performance is artificial neural network in ultrasonographic quantitative diagnosis of hepatic fibrosis and ultrasonic diagnosis earlier period hepatic cirrhosis model on artificial neural network. It is input layer to conventional two-dimensional ultrasound and Doppler ultrasound used in artificial neural network mode, and output layer to the golden standard.Materials and methods1) It were used of that GE LOGIQ 5 Expert and Acuson Sequoia512 color computer ultrasound diagnostic apparatus, and the choice of 3.5C scarching unit, imaging frequency of 5.5MHz, abdominal inspection conditions. 2) It were collected that the ultrasound examination data of liver fibrosis patients and healthy people.Ten ultrasound parameters were been quantization. All data will be selected by the statistical analysis. It was being establishment artificial neural network mode of liver fibrosis diagnostic classification.3) Otherwise, it as drilled in artificial neural network in ultrasonographic grading diagnosis of hepatic fibrosis. The output data and the pathology gold standard established four standard tables. To computation in appraising the artificial neural networks diagnostic model authentic evaluating indicator. Mainly includes: The sensitivity, the peculiarity, the accuracy and so on.4) It sames liver fibrosis graduation diagnosis model building and evaluation, as drilled in ultrasonic diagnosis earlier period hepatic cirrhosis model on artificial neural network.Results1) The liver biopsy confirmed to the liver fibrosis after the ultrasound examination later in chronic liver disease patients. It as training sample that the 81 patients and 22 normal controls of ultrasound imaging data2) That was being evaluated artificial neural network of 65 cases of liver fibrosis cases and 53 cases of normal collected. The sensitivity is 95.4%, and the specificity is 96.2%, and accuracy rating is 95.8% of the artificial neural network for the diagnosis of liver fibrosis grading.3) That was being evaluated artificial neural network of 83 cases of chronic hepatitis. The sensitivity is 84.1%, and the specificity is 84.6%, and accuracy rating is 87.9% of ultrasonic diagnosis earlier period hepatic cirrhosis model on artificial neural network.Conclusion1) This experiment through artificial neural networks' method, manufactured the liver fibrosis diagnostic model and the ultrasound diagnosis earlier period diagnostic model. These have better sensitivity and specificity. It provides the concrete mentality and the method for graduation diagnosis for the liver fibrosis.2) The artificial neural networks have the better sensitivity and the specificity to the ultrasound diagnosis early liver cirrhosis. It will have certain sense to the future in the clinical chronic liver disease patient to degree graduation.