论文部分内容阅读
磁共振(MR)图像提供了大量用于医疗检查的信息。精确鲁棒的脑部MR图像分割、特征提取和分类对于临床诊断肿瘤是非常重要的。提出一种新的基于脑部MR图像的肿瘤诊断方法。首先,通过多阈值分割形态学操作检测图像的畸形区域,然后,提取用于分类的高斯混合模型(GMM)特征,最后,利用决策树分类器对肿瘤图像类型进行分类。整个分类过程分为训练和测试2个阶段,训练阶段提取肿瘤图像和非肿瘤图像不同的特征,在测试阶段基于知识库进行肿瘤和非肿瘤分类。使用准确度、误报率和漏检率3个性能指标对算法进行评估,实验结果表明,分类准确度可达91.18%-94.11%,误报率和漏检率在2.94%-4.41%范围内,可以有助于更好的脑部肿瘤诊断。
Magnetic resonance (MR) images provide a wealth of information for medical exams. Precise and robust brain MR image segmentation, feature extraction and classification are very important for the clinical diagnosis of tumors. A new method of tumor diagnosis based on brain MR images is proposed. Firstly, the deformable region of the image is detected by multi-threshold segmentation morphological operations. Then, the GMM features are extracted. Finally, the classification of tumor images is classified by using the decision tree classifier. The whole classification process is divided into two stages of training and testing. During the training stage, different features of tumor images and non-tumor images are extracted, and based on the knowledge base, tumor classification and non-tumor classification are conducted during the testing phase. The algorithm is evaluated using three performance indexes: accuracy, false alarm rate and missed detection rate. The experimental results show that the classification accuracy can reach 91.18% -94.11%, the false alarm rate and the missed detection rate range from 2.94% -4.41% , Can help better brain tumor diagnosis.