Application And Research Of The Classification Of Liver Fibrosis Based On Medical Images

Liver fibrosis is a common progressive condition in many chronic liver diseases.The early accurate assessment of liver fibrosis can be beneficial for the treatment of mild liver fibrosis and the control of severe liver fibrosis.It also plays a vital role in the management,surveillance and prognosis of patients with liver cancer.However,there are still some difficulties in predicting the degree of hepatic fibrosis.At present,the traditional statistical analysis and machine learning analysis methods are mainly based on the extraction of medical image features.But the current feature extraction methods are time-consuming and labor-intensive.And these features may not appropriately describe the information in images.Otherwise,the acquisition of medical image is difficult and the amount of medical data is relatively small,which would limit the application of deep learning methods.In view of the above problems,this article uses Gd-EOB-DTPA-enhanced hepatocyte-phase MR images to conduct the classification.In this paper,a new Cross-Contrast Neural Network based on symbolic sequence similarity theory is proposed for the classification of hepatic fibrosis.The main contents of this paper are as follows.Firstly,the related background and research significance of liver fibrosis disease and the application of current medical imaging in the diagnosis of hepatic fibrosis are introduced.The basic principles of magnetic resonance imaging and Gd-EOB-DTPA-enhanced MRI are also introduced.The reason of the chosen of Gd-EOB-DTPA-enhanced hepatocyte-phase MR images is also briefly described.Then,a Cross-Contrast Neural Network based on IBS method and transfer learning is proposed.The convolutional neural network is used to extract features from the medical images.The IBS method is subsequently used to measure the similarity of images.And the classification of hepatic fibrosis is conducted based on the measurement.Finally,the author introduces the design of experiments.The traditional statistical analysis and machine learning methods are performed to classify the liver fibrosis with the extraction of texture features.The classification of liver fibrosis by Cross-Contrast Neural Network is also performed.The accuracy of the classification of mild and sever hepatic fibrosis is 93.3%.By comparing the results with traditional statistical analysis and machine learning,the advancement and the feasibility of cross-contrast neural network is verified.