论文部分内容阅读
目的提出一种对数字乳腺影像计算机辅助诊断中可疑密度分割更为有效的分割方法。方法使用基础的边缘分割算子sobel和离散形式的动态轮廓模型对乳腺影像中的可疑密度区域(肿块)进行两步法分割,边缘检测进行带阈值选择的轮廓初步提取,然后采用部分边缘点作为动态轮廓模型的计算点,获得能量收敛的最终轮廓。结果实现对数字乳腺影像库和乳腺体模影像的分割,并对分割轮廓进行与人工分割轮廓的重叠率计算和ROC曲线计算,对算法进行评价。结论最终分割结果有效降低假阳性概率,提高了分割的特异性。
Objective To propose a more efficient method of segmentation of suspicious density segmentation in digital mammography computer aided diagnosis. Methods Based on the edge segmentation operator sobel and the discrete dynamic contour model, the suspicious density area (mass) in breast images was segmented by two steps. The edge detection was used to extract the contour with threshold selection. Then, Calculate the points of the dynamic contour model to obtain the final contour of energy convergence. Results The segmentation of digital mammography and breast phantom images was implemented. The overlap rate and the ROC curve of segmentation contour were compared with artificial segmentation contour to evaluate the algorithm. Conclusion The final segmentation results effectively reduce the false positive probability and improve the segmentation specificity.