独立分量分析在乳腺钼靶X片感兴趣区域提取中的应用

为乳腺癌早期诊断和乳腺X线影像微钙化点计算机辅助检测的前期预处理,本研究提出基于独立分量分析（ICA）的自动提取新算法并且将其应用于乳腺图像感兴趣区域的自动提取.其具体思路是：（1）将乳腺区域图像提取成等大的子图像作为待测乳腺图像感兴趣区域;（2）将ICA应用于乳腺图像感兴趣区域得到基图像;（3）将待识别乳腺图像感兴趣区域在基图像所构成的子空间进行投影求得待测乳腺图像感兴趣区域的特征矢量;（4）用人工神经网络分类方法进行乳腺图像感兴趣区域的模式判别.对临床实际病例的试验结果表明,该方法的检出率为91%,与同类研究检出率相当.本研究方法简单有效,并具有较高的智能性,为ROI的自动提取提供了新的研究思路.

Extracting ROIs in Mammograms Using ICA

In order to preprocess mammograms for diagnosing the early cases of breast cancer and improve the computational efficiency in the computer-aided detection of micro-calcifications in mammograms,this paper presents a new method based on independent component analysis(ICA),which can extract the region of interests(ROI) automatically.The proposed method is based on a four-step procedure:(1) the mammogram was divided into sub-images of the same size as the pending ROI;(2) the base images were obtained by applying ICA to ROI;(3) the feature vector was extracted by projecting the pending ROI to subspace that is composed of the base images;(4) artificial neural network(ANN) was used to recognize ROI.The method was adopted in the processing of the test samples with a true positive rate(TPR) of 91%.The proposed method was easy to use,effective,and with higher intelligence,which provided a new idea for automatic ROI extraction.