Dokuz Eylul University, Department of Computer Engineering, Graduate School of Natural and Applied Sciences, Izmir, Turkey.
Comput Methods Programs Biomed. 2013 May;110(2):150-9. doi: 10.1016/j.cmpb.2012.11.003. Epub 2012 Dec 25.
Many computer aided diagnosis (CAD) systems help radiologist on difficult task of mass detection in a breast mammogram and, besides, they also provide interpretation about detected mass. One of the most crucial information of a mass is its shape and contour, since it provides valuable information about spread ability of a mass. However, accuracy of shape recognition of a mass highly related with the precision of detected mass contours. In this work, we introduce a new segmentation algorithm, breast mass contour segmentation, based on classical seed region growing algorithm to enhance contour of a mass from a given region of interest with ability to adjust threshold value adaptively. The new approach is evaluated over a dataset with 260 masses whose contours are manually annotated by expert radiologists. The performance of the method is evaluated with respect to a set of different evaluation metrics, such as specificity, sensitivity, balanced accuracy, Yassnoff and Hausdorrf error distances. The results obtained from experimentations shows that our method outperforms the other compared methods. All the findings and details of approach are presented in detail.
许多计算机辅助诊断(CAD)系统帮助放射科医生完成乳房 X 光片中的肿块大量检测这一艰巨任务,并且还提供有关检测到的肿块的解释。肿块的最重要信息之一是其形状和轮廓,因为它提供了有关肿块扩散能力的有价值信息。然而,肿块形状识别的准确性与检测到的肿块轮廓的精度高度相关。在这项工作中,我们引入了一种新的分割算法,即基于经典种子区域生长算法的乳腺肿块轮廓分割,该算法具有自适应调整阈值的能力,可增强给定感兴趣区域的肿块轮廓。该新方法在 260 个肿块的数据集上进行了评估,其轮廓由专家放射科医生手动注释。该方法的性能是根据一组不同的评估指标来评估的,例如特异性、敏感性、平衡准确性、Yassnoff 和 Hausdorrf 误差距离。实验结果表明,我们的方法优于其他比较方法。所有的发现和方法细节都详细呈现。