基于可变形模型提取肿瘤区域的方法研究

为了减少计算机分析乳腺肿瘤影像的数据量，提高肿瘤病灶计算机辅助探测和辅助诊断的精确度，降低假阳性和假阴性，乳腺肿瘤区域的提取是至关重要的。基于乳腺恶性肿瘤和有些良性乳腺瘤(如纤维性瘤)的边缘模糊不清和凹凸多变的特性，传统的边缘提取算子和阈值法分割乳腺肿瘤区域很困难，该文提出应用基于可变形模型的梯度矢量流的方法提取乳腺X光片肿瘤区域，用基于亥姆霍兹最小能量损耗原理的一种新的外部能量代替传统方法的外部能量，从而解决了传统方法的可变形轮廓线不能收敛于目标边缘凹陷处的问题，并取得了良好的效果，为进行乳腺X光片肿瘤区域的形状特征提取提供了前提。

Tumor Region Segmentation of Mammogram Based on Deformable Model

Region segmentation of mammogram is vital in order to decrease the amount of data for computer analysis, improve computer-aided detection and diagnosis of tumor and reduce both the false positive and false negative. Traditional boundary extraction operators and threshold method are difficult to be used to extract malign and some benign breast tumor region with vague and irregular boundaries. So the gradient vector flow based on deformable contour is presented in this paper to segment tumor region of mammogram utilizing a novel external energy in terms of Helmholtz principle of minimum energy dissipation instead of traditional external energy. The presented method addresses the problem of convergence to concavity that cannot be overcame by traditional deformable model and achieves good results. This work provides the precondition for shape features extraction of mammogram.