乳腺影像案例不仅具有图像的底层特征,同时也有图像的语义特征.为了实现乳腺影像的高效检索,提高计算机辅助诊断的确信度,提出了一种基于关联规则的多模检索方法.首先,采用基于关联规则的特征选择算法选择出与影像语义相关的底层特征,实现特征降维,利用Apriofi算法挖掘被选择的特征与语义特征之间的关联规则.然后,利用关联分类引擎算法根据得到的关联规则构建关联分类模型,实现由底层特征获知视觉语义特征的目的.最后,将关联分类模型得到的语义特征作为输入语义,与图像的底层特征相结合,进行图像相似性度量,实现多模检索.通过查准率和查全率以及相关排序平均值等进行了实验对比,实验结果表明,提出的多模检索方法有效的提高了图像的检索精度并且能够由图像的底层特征获知图像的视觉语义特征.该方法缩减了底层特征和视觉语义特征之间的语义鸿沟,提高了图像的检索性能,能够为医生提供更有意义的决策支持.%The mammogram case has images of low level features and semantic features.In order to achieve efficient retrieval of breast imaging cases,and enhance the certainty of computer aided diagnosis,a multi-mode retrieval method based on association rules is proposed in this paper.First of all,feature selection algorithm based on the association rules can be used to select the low level features associated with image semantic features,to achieve the dimension reduction.The associative rules which between the selected features and the semantic features can be excavated by using the Apriori algorithm.And then,the associative classifier engine will be used to build the associative classification model depend on the associative rules to capture the visual semantic features.Finally,take obtained semantic from the association classification as input semantic,combining with the low level features of image,to implement the mammogram case multi-mode retrieval.We conducted experiments comparing by precision and recall rate and relevance ranking average value and so on,as the results show,multi-mode retrieval method proposed by this paper can effectively improve the performance of breast imaging case retrieval,and provide visual semantic features of image by its low-level features.Multi-mode retrieval reduced the semantic gap between image low level features and visual semantic features,improved the accuracy of image retrieval and provided more meaningful decision support for doctors.
展开▼
