Abstract:

Objective  To distinguish hepatocellular carcinoma, liver cirrhosis and normal liver tissues using ³¹P MRS.  Methods  Using a 1.5T superconducting MR scanner, ³¹P MRSdata of 66 effective voxel samples were collected from 33 cases of hepatocellular carcinoma, liver cirrhosis and normal liver volunteers.  The BP neural network and SVM model experiments were used to differentiate hepatocellular carcinoma, liver cirrhosis and normal liver tissues.  Results Limited samples achieved a good classification performance. Back-propagation neural network (BP)and Support Vector Machine (SVM) models can improve identification rates of ³¹P MRSto 92.31%.  Conclusion  ³¹P MRSdata analysis based on BP and SVM for in vivo diagnosis of hepatocellular carcinoma provides a valuable optional technology.



Key words: Carcinoma， hepatocellular; liver cirrhosis; magnetic resonance spectroscopy；    Support vector machine; Back-propagation  neural network