基于MBD的乳腺超声图像肿瘤分割方法

A Tumor Segmentation Method of Breast Ultrasound Image Based on MBD

计算机辅助诊断(CAD)系统越来越多的应用于临床乳腺超声检查中,乳腺超声图像肿瘤分割在CAD系统中扮演着关键角色,决定最终的分析质量。现有分割方法在图像质量、先验知识、数据量等方面有着一定的要求和限制条件,适应性存在很大的局限。根据乳腺超声图像的特点,参考视觉注意力机制,将基于最小障碍物距离的显著性检测算法应用于乳腺肿瘤的分割任务中。实验表明,该方法能够更为精确地分割乳腺超声图像中的肿瘤部分,即使在超声图像质量很差的情况下也能达到很好的分割效果,具有更好的适应性。

Computer Aided Diagnosis(CAD)system is more and more used in clinical breast ultrasound examination.Tumor segmentation of breast ultrasound image plays a key role in the CAD system,which determines the final analysis quality.The existing segmentation methods have certain requirements and limitations in image quality,prior knowledge,data volume and so on.According to the characteristics of breast ul⁃trasound image and the mechanism of visual attention,the saliency detection algorithm based on the minimum barrier distance is applied to the segmentation of breast tumor.Experimental results show that this method can segment the tumor part of breast ultrasound image more accurately,even in the case of poor ultrasound image quality,it can achieve good segmentation result,and has better adaptability.