基于乳腺钼靶侧斜位图像的近期乳腺癌风险预测研究  

Prediction of Near-Term Breast Cancer Risk in Mediolateral Oblique View of Mammography

作　　者：李颜娥 张朋 范明 Zheng Bin 厉力华 Li Yane;Zhang Peng;Fan Ming;Zheng Bin;Li Lihua(College of Life Information Science and Instrument Engineering, Hangzhou Dianzi University, Hangzhou 310018, China;School of Electrical and Computer Engineering, University of Oklahoma, Oklahoma, Tulsa, OK 74135, USA)

机构地区：[1]杭州电子科技大学生命信息与仪器工程学院,杭州310018 [2]School of Electrical and Computer Engineering,University of Oklahoma

出　　处：《中国生物医学工程学报》2018年第2期169-180,共12页Chinese Journal of Biomedical Engineering

基　　金：国家自然科学基金(61271063;61401131);浙江省重点基金(LZ15F01001);国家重点基础研究发展计划项目(2013CB329502);Grants R01CA197150 from the National Cancer Institute;National Institutes of Health;USA

摘　　要：探讨乳腺钼靶侧斜位图像双侧不对称性与近期乳腺癌风险之间的关联性。回顾性分析556例连续两年进行乳腺普查的样本,当前年患癌与阴性样本各278例,且两组样本年龄匹配。对前一年乳腺钼靶侧斜位图像进行预处理获取双侧局部对应区域及全局区域后分别提取空间差异性、结构相似性及位置信息等85维影像特征。删除相关性较高的特征后剩余78个特征,采用留一法与逐步回归分析进行特征选择,并建立基于广义线性模型的近期乳腺癌风险预测模型。结果显示,全局及局部不对称性特征相结合进行风险预测AUC值为0.666 7±0.022 6,对其进行混淆矩阵分析,特异性为0.690 6,灵敏度为0.521 6。对风险预测值进行回归分析,优势比(Odds Ratio)随风险的增加而显著增加,P值为0.002 033。3个基于年龄的子集(37~49岁,50~65岁,66~87岁)进行近期乳腺癌风险预测,AUC值分别为0.681 0±0.043 2,0.671 6±0.030 0及0.678 2±0.054 7,对应3个年龄段的特异性分别为0.702 7,0.694 3和0.723 4,灵敏度分别为0.554 1,0.490 4和0.574 5。两个基于乳腺X线分型的子集(BIRADS 2、BIRADS 3)AUC值分别为0.654 5±0.036 9及0.694 4±0.03,特异性分别为0.676 2和0.733 3,灵敏度分别为0.522 9和0.536 9。结果表明,乳腺钼靶侧斜位图像全局及局部不对称性相结合对近期乳腺癌风险预测具有潜在有效价值。This study proposed a model for the prediction of near-term breast cancer risk in mediolateral oblique view（ MLO） of negative full-field digital mammography（ FFDM） images based on local and global region bilateral asymmetry features. The retrospective dataset included series of two sequential FFDM examinations of556 women. In the ＂current＂examination 278 women were diagnosed with pathology verified cancers and 278 cases remained negative. Patients ＇age for breast cancers and negative cases were matched. After region of interest include local and global regions were segmented,spatial distribution,structural similarity and positional information related features were extracted. After decorrelation process,7 features with Spearman correlations 0. 6 were excluded and 78 features were remained for further analyses. Next,a short-term breast cancer risk prediction model was built using a leave-one-case-out cross-validation method to predict the likelihood of each woman having image-detectable cancer in the next sequential mammography screening. The computed areas under a receiver operating characteristic curves（ AUC） was 0. 6667±0. 0226 with the specificity and sensitivitywere 0. 6906 and 0. 5216 respectively when combined global-and local-based features. The odds ratio values was increased with a significantly increasing trend in slope（ P = 0. 002033） as the model-generated risk score increased. In addition,for the three age groups of 37-49,50-65 and 66-87 years old,the AUC values were0. 681 0±0. 043 2,0. 671 6±0. 030 0 and 0. 678 2 ± 0. 054 7 with the specificity were 0. 702 7,0. 694 3 and0. 723 4 and sensitivity were 0. 554 1,0. 490 4 and 0. 574 5 respectively. And AUC values of 0. 654 5 and0. 694 4 were yielded for BIRADS 2 and BIRADS 3,respectively. With the specificity were 0. 676 2 and0. 733 3 and sensitivity were 0. 522 9 and 0. 536 9 for BIRADS 2 and BIRADS 3,respectively. This study demonstrated the potential of bilateral asymmetry features extracted from MLO view mammography to assis

关 键 词：双侧乳腺钼靶侧斜位图像 近期乳腺癌风险 双侧不对称性 局部区域 

分 类 号：R318[医药卫生—生物医学工程]