Research On Several Key Techniques Of Hepatic Computer-Aided Diagnosis Based On MRI

Our country has already been in the worst-hit areas of liver disease, the mortality of liver cancer ranks the second, and has been aroused more attention. The deveopment of Computer-Aided Diagnosis (CAD) is critical to improve the accuracy of diagnosing liver diseases, reduce the error ratio of diagnosis and improve work efficiency. The hepatic CAD system is made up by several modules, such as image preprocessing, hepatic area extraction, feature selection, classification recognition and 3D visualization. This paper mainly conducts on the pretreatment of the abdominal Magnetic Resonance Imaging (MRI), the extraction of hepatic target area,3D visualization and feature extraction. The main content of this paper is shown in the following paragraphs.Firstly, there is existing a strong intensity inhomogeneity among the abdominal MRI, which causes overlapping of intensity in different tissues, and it is difficult to obtain the accurate and fast segmentation result. This paper has been presented the Mask-Weight Fuzzy Clustering Method (M-WFCM) algorithm that incorporates the local intensity information, global consistency based on intensity information and spatial information, energy constraints based on the bias field, to accurately extract intensity inhomogeneity among abnirmal MRI based on the smoothness of bias field in this paper. Our experiment has been applied to the real abnormal MRI that hospital providing, the results show that M-WFCM algorithm has high versatility, and the histogram of corrected images appears obvious peaks and valleies, which means the corrected image has been achieved the desired result.Secondly, the complexity of the hepatic structure will increase the precise segmentation of liver tissue difficulty. In order to solve this problem, this paper conducts the liver extraction based on the intensity inhomogeneity correction. First of all, it applies the threshold segmentation algorithm based on the histogram information and morphological operations to obtain the initial contour of liver. Then, this paper uses the modified Level set algorithm to accurately segmente based on the initial contour.Then, this paper applies the 3D reconstructed algorithms to reconstruct and display for hepatic tissue. This paper mainly extracts liver based on different forms, and the surface rendering technology of hepatic tissue provides the external contour information. This kind of algorithm has characteristics of fast operation, the algorithm principle is simple, and can achieve real-time operation. Besides, this paper applies Maximum Intensity Projection (MIP) algorithm to reconstruct the hepatic tissue for providing the blood vessels and texture information in hepatic tissues. The experimental results show that MIP algorithm of volume rendering can provide more detail information of liver tissue, and has a high value of practical application.At last, feature selection is an important module of liver CAD, and the good or bad of feature selection will directly affect the classifier’s generalization ability and the accuracy of classification. This paper mianly study the network structure optimization based on Constructive Approach for Feature Selection (CAFS) algorithm with feature selection.