基于小波变换的乳腺X线影像微钙化点感兴趣区域提取新技术  

A Novel ROI Extracting Technique Based on Wavelet Transform for the Detection of Micro-calcifications in Mammograms

作　　者：李树楠[1] 万柏坤[1] 马振鹤[1] 王瑞平[1] 

机构地区：[1]天津大学生物医学工程系,天津300072

出　　处：《生物医学工程学杂志》2005年第2期360-362,共3页Journal of Biomedical Engineering

摘　　要：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作前期预处理,提出了一种基于小波变换的微钙化点感兴趣区提取新技术。其具体思路是:(1)将乳腺区域图像提取成等大的子图像;(2 )对每一幅子图像进行小波变换,根据特征参数ρ讨论最优小波变化参数和阈值T;(3)根据阈值T判别子图像是否属于感兴趣区。对临床实际病例(2 0幅乳腺X线影像)的试验结果表明,该方法具有较高的检出率(89.7% ) ,和较为满意的假阳性率(2 .1% )。In order to preprocess mammograms for diagnosing the early cases of breast cancer and improving the computational efficiency in the computer-aided detection of micro-calcifications in mammograms, we have advanced a novel processing technique for the extraction of micro-calcificatlon region of interest （MROI）. The proposed method is based on a three-step procedure： （1） the mammogram is divided into sub-images of the same size; （2） the wavelet multi-resolution method is conducted on the sub-images, and the parameters related to wavelet transform and threshold T are discussed according to ρ; （3） the classification of sub-images is determined by T. It is tested with 20 mammograms and the results show that the method can achieve a true positive rate as high as 89.7% with a false positive rate as low as 2.1%.

关 键 词：乳腺X线影像 微钙化点感兴趣区域(MROI) 图像分割 小波变换 乳腺癌早期诊断 感兴趣区域提取 微钙化点 X线影像 计算机辅助检测 图像提取 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]