【摘要】 提出一种自适应的肺结节检测算法。基于肺结节的CT图像表现,提取了七个图像特征。对于每个特征,采用支持向量机(SVM)进行单特征分类,将其分类的准确率作为权值应用到改进的Mahalanobis距离公式中。根据改进的Mahalanobis距离公式构造非线性分类器,然后对感兴趣区域(ROI)进行分类,分类器中使用自适应的算法自动调节阈值。实验结果表明,该算法对于肺结节的检测具有较高的敏感性和准确性。更多还原

【Abstract】 An adaptive detection algorithm for pulmonary nodules was proposed. Based on CT images of pulmonary nodules, seven image features were extracted. Then Support Vector Machine (SVM) algorithm was applied to each feature vector to classify the pulmonary nodules, and the accuracy of each feature classification was used as a weight in the modified Mahalanobis distance formula. A nonlinear classifier, which classified the regions of interest (ROI), was constructed on the base of modified Mahalanobis distance formula. The adaptive algorithm was used to adjust the threshold in the classifier. The experiment indicates that the algorithm has a good sensitivity and accuracy for pulmonary nodule detection.更多还原