Hepatic Pathological Changes Computer Aided Diagnostic Based On MRI

Liver cirrhosis is a common clinical chronic liver disease with the diffuse liver damage due to the repeat action of one or more reasons. Hepatic fibrosis is a pathological process that an abnormal liver connective tissue hyperplasia because of some kind of pathogenic factor make the diffuse liver extracellular matrix deposit too much, which will lead to cirrhosis, liver fibrosis is an important link to the development of cirrhosis and chronic liver diseases. Currently, the diagnostic methods of liver cirrhosis and hepatic fibrosis include liver biopsy, blood biochemical examination, ultrasound, computed tomography (CT), magnetic resonance imaging (MRI) and so on. As an invasive diagnostic method, liver biopsy is the "gold standard ", but it will misdiagnose due to the uneven distribution of hepatic pathological changes and the limited liver biopsy scope. Therefore, it has practical significance to explore a high-performance and non-invasive diagnostic method instead of liver puncture. In the framework of pattern recognition, computer-aided diagnosis (CAD) technology based on MRI fully excavate the information in the image based on medical image processing and analysis and combine with clinical information, obtain more accurate and objective diagnosis. This paper study the liver cirrhosis and hepatic fibrosis CAD based on MRI systematically and propose three methods including human liver cirrhosis CAD based on multiple sequences MRI, the classification of human liver cirrhosis based on random projection texture features and rat hepatic fibrosis CAD based on equilibrium phase of enhanced MRI. The main work of this paper is as followed:(1) Human liver cirrhosis CAD based on multiple sequences MRI:in the framework of digital image processing and pattern recognition, according to the region of interest (ROI) process, classification of multiple MRI of ROI and the individual classification, this method complete the human cirrhosis CAD with the help of neural networks, decision trees voting mechanism and complete human cirrhosis. This method combines the multi-parameters, multi-sequences, multi-dimensional imaging and it has many advantages such as information-rich, structured, high accuracy and so on.(2) The classification of human liver cirrhosis based on random projection texture features:this method classify the samples into normal, early liver cirrhosis and advanced liver cirrhosis with the texture features based on random projection. Experiments show that this method is better than the classic texture features based on gray level co-occurrence matrix (GLCM) of neural network models in complexity, computational complexity, time consuming, and accuracy.(3) Rat hepatic fibrosis CAD based on equilibrium phase of enhanced MRI:this method classifies the rat hepatic fibrosis with the texture features based on the GLCM which extracting the non-linear quantization ROI and neural network. Experiment shows that the non-linear quantization texture features have a positive effect in rat hepatic fibrosis classification.