Design And Implementation Of Pancreatic Classification System Based On Sparse Representation

Pancreas as the second large glands next to liver in the body has important significance for the digestion of food,Therefore the diagnosis of pancreatic diseases is also of great significance to the health of human body.CT image is one of the momentous methods to diagnosis of pancreas diseases in early stage with the advantage of high resolution and little damage to human bodies and the accurate response to the pathological change of pancreas.However,the doctor’s experience in reading graphs and subjecting factors have a significant influence on the diagnosis results of pancreatitis in anterior period.Especially in the presence of chronic pancreatitis cases,the diagnosis result of pancreatic cancer is often subject to interference,even for the experienced doctors may draw a false conclusion.At the same time,considerable CT images add great workload to doctor.Consequently,it’s urgent for medical field to strengthen the auxiliary diagnosis technology of computer on pancreatic cancer image.In this context,based on the deep study for the feature extraction,feature selection and sparse representation classifier key algorithm,the pancreatic classification system based on sparse representation theory was developed.This system mainly composed of five parts that is selection of the interested regions,feature extraction of pancreas,feature selection of pancreas,construction the over.complete redundant dictionary and pancreatic classifierbased on sparse representation.At first,the interested regions need to be picked out from the CT image,then the four feature extraction methods were adopted to gain the accurate information to describe the features of pancreas,meanwhile,the redundant featured information can be eliminated through feature selection;afterwards the over.complete redundant dictionary would be constructed based on these features,at last,the pancreatic classifier model was devised based on sparse representation which can be used to identify the pancreas images.In order to testify the performance of the system,the pancreatic recognition experiment via CT image based on the methods mentioned above was carried out.The results showed that the method was feasible,effective in feature selection and accurate in the classification system based on sparse representation.Compared with the traditional SVM,this method is more superior and the identification result is also satisfactory which can provide doctors with a reference for the auxiliary diagnosis.