多视图信息融合的乳腺肿块计算机辅助检测关键技术研究

项目名称： 多视图信息融合的乳腺肿块计算机辅助检测关键技术研究

项目编号： No.61302192

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 无线电电子学、电信技术

项目作者： 徐胜舟

作者单位： 中南民族大学

项目金额： 25万元

中文摘要： 乳腺癌是中老年妇女主要死因之一，基于乳腺X线摄片的计算机辅助检测系统可有效减轻乳腺肿块诊断过程中人工阅片的工作强度并提高准确率。但由于肿块大小各异、边缘模糊，常被致密组织所掩盖，且其在单幅乳腺X线射片中的信息有限，目前基于单视图的检测系统普遍存在假阳性率偏高的问题。本项目通过模仿放射科医师的阅片机制提出一种多视图信息融合的检测方法来提高检测性能。首先，利用提出的基于模板匹配的多同心层检测算法对乳腺的轴位和斜侧位视图分别进行肿块检测；然后，采用基于动态规划和形态学的分割算法来提取肿块轮廓及毛刺；再设计一种乳房全局配准和可疑肿块区域局部配准相结合的方法获取可疑肿块区域在对侧乳腺相同视图上的对称区域并与之组成区域对，融合区域对的特征信息来识别假阳性区域；最后，融合该乳腺的轴位和斜侧位视图上的可疑肿块的特征信息以进一步去掉假阳性区域。该研究将提高乳腺肿块计算机辅助检测系统性能，推动其临床普及应用。

中文关键词： 乳腺X线射片；肿块；多视图；检测；

英文摘要： Breast cancer is one of the leading cause in mortality among middle-aged and older women. Computer-aided detection(CAD)systems based on mammograms have been developed to improve the performance of breast mass diagnosis by radiologist. However, due to the various sizes,blur edges, overshadowing by the surrounding dense tissues, and the limited information in the single mammograms, the false positive rate of current computer-aided detection of mass based on single image remains relatively high. To solve this problem, a computer-aided detection method of breast mass based on information fusion of multi-view is proposed by imitation the radiologist's manual screening mechanism, in this project. Firstly, a method based on multiple concentric layer and template matching is designed to detect masses in the cranio-caudal view and the medial lateral oblique view of the mammogram, respectively.Secondly, the central mass border and spicules are extracted by a segmentation method based on dynamic programming and morphological method. Thirdly, for each suspicious region detected, the symmetric region on the same view of the contralateral breast is identified by combining the global registration of breast region and local registration of mass region. The suspicious regions and corresponding symmetric regions are matched and t

英文关键词： Mammograms；Mass；Multi-view；Detection；