Optimal Feature Selection and Classification of Liver Tumors in Computed Tomography Images

• Main Authors: Ya-Han Chang, 張雅涵
• Other Authors: 郭文嘉
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/38218997976940991718

碩士 === 元智大學 === 資訊管理學系 === 98 === In this study, we propose to a computer-aided diagnosis (CAD) system for the classification of liver tumors in non-enhanced computed tomography (CT) images. This study aimed to classify liver tumors to assist doctors for further diagnosis. The proposed system consists of three parts. First, the feature extraction module extracts 102 statistical and texture features. Second, the feature selection module acquired the combination of the best features by integrating the particle swarm optimization (PSO) with support vector machine (SVM) to reduce the complexity of computation. Finally, a support vector machine based classification model was constructed to identify benign and malignant liver tumors. Experimental results show that the accuracy of the proposed CAD system for classifying liver tumors was 84.53%, the sensitivity was 80%, the specificity was 88.09%, the positive predictive value (PPV) was 88.77% and negative predictive value (NPV) was 84.01%.The accurate rate is up to 80 % both on benign and malignant tumors of the CT images. The proposed method can achieve the purpose of enhancing the accuracy of automatic identify effectively to assist further diagnosis.