【摘要】 深度学习技术逐渐成为自动检测乳腺X线图像肿块的主流技术,然而由于肿块的大小、形状、边界、纹理存在多样性,并且肿块的信噪比比周围组织低,目前的算法仍然存在检测遗漏目标的问题.为了有效地提高检测精度,提出了一种基于空间约束和多特征融合的多级目标检测架构,命名为SC-FU-CS RCNN.首先,该框架由多个递增的重叠度(IoU)阈值训练的检测器构成,更深层的检测器可以更好地减少误报;其次,在框架中融合了卷积神经网络(CNN)的浅层与深层特征,使检测器更好地检测图像中的模糊、较小的目标;最后,添加空间约束层纳入目标周围的拓扑区域特征.实验表明,该框架在乳腺X线检查数字数据库(DDSM)平均准确率达到94.06%,优于其他相关算法.更多还原

【Abstract】 In recent years,deep learning technology has gradually become the mainstream technology for automatic detection of mammograms.Currently,object detection algorithms do not eliminate the problem of missing cancer,because the shape,size,boundary and texture of the masses are variety and also because the signal to noise ratio of the masses is lower compared to the surrounding breast tissue.In this paper,we present a novel approach for detecting masses in mammograms using a multi-stage object detection architecture with spatial constraints and multi-feature fusion,which is named SC-FUCS RCNN.First,it consists of a sequence of detectors trained with increasing IoU thresholds,to be sequentially more selective against close false positives.Second,we improve the accuracy of the detection by concatenating the shallow and deep layers of the CNN,the detector can detect blurrier or smaller masses in mammograms.Finally,we add a spatial constrained layer before the output layer.Our results show that the proposed method achieves an overall accuracy of 94.06% on the mammography digital database(DDSM),which is superior to other related algorithms.更多还原