Feature Selection In MRI Based Bladder Cancer CAD System

Computer Aided Diagnosis(CAD for short), is a popular technology produced by the combination of a modern high-definition imaging technology and computer analytical calculation.It has considerable value in clinical applications. The entire CAD system consists of four steps:ROI (region of interest) extraction, feature extraction,feature selection and classification. Asmore and more features have been put forward, feature set is grownlarger and includes a large number of duplicate information.It may even lead to the "curse of dimensionality". Therefore, a feature selection algorithm which can make an effective feature selection is urgently needed in the CAD system. In this paper, the author does some research on the feature selection algorithms in CAD system:Firstly, this paper discusses the current situation and problems in feature selection, proposes the basis and techniques of selecting model.Thenit analyzes some commonly used feature selection algorithm of Filter model through some convincing experiments.Secondly, the author proposes an algorithm called mR-FAST, in order to improve the selection effect. Experiments prove that this algorithm has achieved the desired effect. The Experiments use the ROC curve (Receiver Operating Characteristic Curve) as the standard of classification effect, which makes the evaluation more in line with the significance of medical research.Finally, the author applies mR-FAST algorithm to MRI based bladder cancer computer-aided diagnosis systemand and finally the accuracy achieves98.3713%, the AUC achieves0.9852.