| 乳腺肿瘤超声图像形态特征选择 |
| |
| 引用本文： | 林江莉,汪天富,彭玉兰,蒋银宝.乳腺肿瘤超声图像形态特征选择[J].四川师范大学学报(自然科学版),2005,28(5):615-618. |
| |
| 作者姓名： | 林江莉 汪天富 彭玉兰 蒋银宝 |
| |
| 作者单位： | 四川大学,生物医学工程系,四川,成都,610065;四川大学华西医院,超声科,四川,成都,610041 |
| |
| 基金项目： | 四川省青年科技基金和四川省应用基础研究基金资助项目 |
| |
| 摘 要： | 形态特征是超声诊断乳腺良、恶性肿瘤的重要依据.拟建立最佳形态特征矢量,研究提取了似圆度、平均标准化径向长度、熵等7种形态学特征,再用Logistic回归模型对特征进行选择,最终选取似圆度和面积比率这两个特征量组成最佳特征矢量.对经病理证实的乳腺良、恶性肿瘤超声图像进行识别,恶性肿瘤的识别率为93.3%,误判率为12.5%,良性肿瘤的识别率为88.2%,误判率为6.25%.
|
| 关 键 词： | 乳腺肿瘤 超声图像 形态特征 Logistic回归模型 |
| 文章编号： | 1001-8395（2005）05-0615-04 |
| 收稿时间： | 2005-05-25 |
| 修稿时间： | 2005年5月25日 |
Selection for Morphologic Features in Ultrasonic Image of Breast Tumor |
| |
| LIN Jiang-li,WANG Tian-fu,PENG Yu-lan,JIANG Yin-bao.Selection for Morphologic Features in Ultrasonic Image of Breast Tumor[J].Journal of Sichuan Normal University(Natural Science),2005,28(5):615-618. |
| |
| Authors: | LIN Jiang-li WANG Tian-fu PENG Yu-lan JIANG Yin-bao |
| |
| Abstract: | Morphologic features are an important basis to diagnose breast tumors by B-scan ultrasound. In this paper, the optimum morphologic feature vector was constructed to classify breast tumors as benign or malignant. Seven morphologic features, such as pseudo-circularity, mean normalized radial length and entropy, were extracted in this research. And the Logistic regression model was then used for the features selection. Finally, these two features, pseudo-circularity and area ratio, were selected and combined to form an optimum morphologic feature vector. By using a database of breast ultrasonic images, the result shows that the positive predictive value is 93.3%, positive error rate is 12.5%, and the negative predictive value is 88.2%, negative error rate is 6.25%. |
| |
| Keywords: | Breast tumor Ultrasonic image Morphologic features Logistic regression model |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
