多分辨率特征提取方法用于检测乳腺透视图象中的星形肿块

本文提出一种检测乳腺透视图象中星形肿块的多分辨率特征提取方法.首先使用线性相位、不可分小波变换得到原始乳腺透视图象的多分辨率表示;其次选择能将星形肿块从正常乳腺结构中区别开的4个特征,在每种分辨率上计算每个象素的特征.这一方法解决了需获取不同大小的目标但却无法预先选择确定大小邻域的难题.文中也简要介绍了根据这一特征提取方法得到的特征集来使用神经网络对象素分类的过程.最后给出了对于测试样本的特征提取结果。

Multi-Resolution Feature Extraction of Spiculated Lesions in Digital Mammograms

In this paper we present a novel multiresolution feature extraction scheme for the detection of spiculated in digital mammogram. First, a multiresolution representation of the original mammogram is obtained using a linear phase nonseparable wavelet transform. 5 features that can differentiate spiculated lesions from normal structure of mammogram are chosen. This approach addresses the difficulty of predeterming the neighborhood size for feature exaction to characterize objects that may appear in different sizes. The procedure that classifies pixels with neural network according obtained feature set is also briefly described. At last the result of feature extraction for a test pattern is presented.