Design And Implementation Of Medical Image Mining System Based On PMML

With the rapid progress of digital image industry, medical information databases store not only large amounts of structured data of patients, but also masses of unstructured medical image data. These data provide rich data resources for Image Data Mining. Medical Image Mining System is used to mine efficient rules and models, so as to accelerate the process of diagnosis and improve the diagnostic accuracy.This paper focuses on the main techniques of data mining on medical images. A new data mining language CMQL is proposed, and a medical image mining system based on CMQL and PMML (Predictive Model Markup Language) is developed. The research results can be used in the computer-aided diagnosis system. The major achievements in this paper can be generalized into the following aspects.1. On the basis of the analysis of medical image characteristics,information model and the DICOM image file structure, a tool has been designed and developed with Java, which can extract the information from DICOM image file and display the DICOM image.2. The definition and grammar of PMML had been analyzed as well as data flow. On the basis of these, we develop a PMML model conversion tool. This tool supports both the output and the import of PMML files, so as to overcome the shortcoming of most data mining systems, which only provide the output function of PMML files.3. On the basis of analyzing the existing data mining languages, we propose a new data mining language CMQL. CMQL is mainly used for the definition of mining tasks.4. We defined a CTM file format for data mining. This file provides the unified view for the data mining system. It is convenient to preprocess and improve mining rate. This file avoids the abnormal operation which most system mining databases will arose.5. On the basis of analyzing the Java Component Development Technology, we design and develop a data mining prototype system for medical image. It supports CMQL and PMML, and provides the function of mining algorithm component. This system uses the Hashmap model to manage data mining algorithms, which facilitates the search,update and verification of mining algorithms.