Multi-classifier Combination And Its Application In Medical Image Classification

Medical image has already been an important clinic diagnosis basis for disease diagnosis, decision-making before surgery and random visiting after surgery. Diagnosis by medical image is one of the main methods of none wound diagnosis. Research on medical image diagnosis is an important research direction of medical area both at home and abroad. It has high academic values and broad application foreground by applying classification of data mining and computer technology to analyze, compute and process medical images, so as to mine abundant feature information and rules which can aid doctors to diagnose.Currently, the research on medical image classification technology has just started, and there are a lot of problems by applying existing classification methods directly to medical images. Investigating and researching some classification methods and algorithms appropriate for medical images have some quite important practical values.Destination, significance, abroad and domestic research status of medical image classification are discussed, and medical image classification methods fitting for medical image feature as well as its application are analyzed. In order to overcome the problems of current classification algorithms, mining algorithms based on multi-classifier combination are applied, and considering that when mining data of the continuously renewed huge image database algorithms in existence are low in efficiency etc, the incremental classification is imported in and a basic medical image classification system based on multi-classifier combination is built.This paper takes medical image data as research objects, and make a study on medical image classification from three facts, including theory, algorithm and application. Research fruits include following contents: 1.The medical image feature extraction is summarized completely, and the extraction methods of gray histogram features, gray co-occurrence matrix features and wavelet features are analyzed and used to extract features of liver CT images, then several typical classification algorithms are analyzed and realized.2. Detailedly studies Cascade combined model, then in order to break through the accuracy limit of single classifiers, two combined classifiers based on Cascade model are designed and applied to medical image classification, their performance are analyzed and compared.3. In order to overcome disadvantages of combined classifiers that consume so much system resource because of the complicated classification process, an incremental classification model based on Cascade combination is proposed, then the single classifiers, the non-incremental and incremental combined classifiers are realized and compared in their performance.4. Research on the medical image classification model based on combined classifiers, primaryly analyses the extraction method of classification rule of combined classifiers, finally give the description of liver images' classification rule.