Reljin Branimir, Milosević Zorica, Stojić Tomislav, Reljin Irini
Faculty of Electrical Engineering, University of Belgrade, Belgrade, Serbia.
Folia Histochem Cytobiol. 2009 Jan;47(3):525-32. doi: 10.2478/v10042-009-0076-1.
Two methods for segmentation and visualization of microcalcifications in digital or digitized mammograms are described. First method is based on modern mathematical morphology, while the second one uses the multifractal approach. In the first method, by using an appropriate combination of some morphological operations, high local contrast enhancement, followed by significant suppression of background tissue, irrespective of its radiology density, is obtained. By iterative procedure, this method highly emphasizes only small bright details, possible microcalcifications. In a multifractal approach, from initial mammogram image, a corresponding multifractal "images" are created, from which a radiologist has a freedom to change the level of segmentation. An appropriate user friendly computer aided visualization (CAV) system with embedded two methods is realized. The interactive approach enables the physician to control the level and the quality of segmentation. Suggested methods were tested through mammograms from MIAS database as a gold standard, and from clinical praxis, using digitized films and digital images from full field digital mammograph.
描述了两种用于数字或数字化乳腺X线摄影中微钙化分割和可视化的方法。第一种方法基于现代数学形态学，而第二种方法使用多重分形方法。在第一种方法中，通过使用一些形态学操作的适当组合，可以获得高局部对比度增强，随后显著抑制背景组织，而不管其放射密度如何。通过迭代过程，该方法仅高度强调小的明亮细节，即可能的微钙化。在多重分形方法中，从初始乳腺X线摄影图像创建相应的多重分形“图像”，放射科医生可以自由改变分割水平。实现了一个嵌入两种方法的合适的用户友好型计算机辅助可视化（CAV）系统。交互式方法使医生能够控制分割水平和质量。建议的方法通过来自MIAS数据库的乳腺X线摄影作为金标准进行了测试，并在临床实践中使用了来自全场数字化乳腺X线摄影的数字化胶片和数字图像进行了测试。
