Halkiotis Stelios, Mantas John
Health Informatics Laboratory, University of Athens-Faculty of Nursing.
Stud Health Technol Inform. 2002;90:24-9.
In this paper we propose a new algorithm for the detection of clustered microcalcifications using mathematical morphology and artificial neural networks. Considering each mammogram as a topographic representation, each microcalcification appears as elevation constituting a regional maxima. Morphological filters are applied, in order to remove noise and regional maxima that doesn't correspond to calcifications. Each suspicious object is marked using a binary image and finally a feed forward neural network classifies every object achieving a rate of 90% true positive detections with 0.11 false positives per image.
在本文中，我们提出了一种利用数学形态学和人工神经网络检测簇状微钙化的新算法。将每幅乳房X光照片视为一种地形表示，每个微钙化表现为构成区域最大值的隆起。应用形态学滤波器，以去除噪声和与钙化不对应的区域最大值。使用二值图像标记每个可疑对象，最后，前馈神经网络对每个对象进行分类，实现了90%的真阳性检测率，每张图像的假阳性率为0.11。
