基于视觉注意模型的肝脏病灶区域自动提取方法

Auto-detection algorithm of liver focus based on visual attention model

医学图像的感兴趣区域(region of interest,ROI)提取在计算机辅助诊断中有着重要作用。本文借助视觉注意模型的框架,针对肝脏区域病变组织与周围组织灰度差异小、弱纹理的特点,提出了一种新的肝脏CT(computer tomography)图像中异常组织区域提取方法。本方法首先筛选一组与肝脏纹理相关的统计特征和基于方向分形维的显著度特征因子,然后由各特征分量显著性图合成总显著性图,最后根据显著性图标识病变发生的区域。试验表明,本文方法能准确找到肝脏区域中发生病变的位置,为显著性图在医学影像ROI提取提供了有效途径。

The detection of region of interest（ROI） in medical images has played a very important role in computer aided diagnosis.Because liver-focus organs have the characteristics of weak textural and small intensity differences with their neighborhood,a novel algorithm of extracting abnormal regions in liver CT images has been proposed in this paper,which uses visual attention model.Firstly,a set of statistical features related to liver textures and some salient factors based on directional fractals are selected.Then,an overall saliency map is composed from several subsalient maps of feature components.Finally,the regions with liver focus are located by labeling the saliency map. Experiments show that the locations of liver focus regions could be found using the proposed method accurately and using saliency maps is an effective way of extracting ROI in medical images.