基于多参数MRI的前列腺癌计算机辅助检测方法
Computer-aided Detection of Prostate Cancer based on Multiparametric MRI
投稿时间:2018-12-23  修订日期:2018-12-23
DOI:
中文关键词: 计算机辅助检测  前列腺癌  多参数磁共振成像
英文关键词: computer-aided detection  prostate cancer  multiparametric MRI
基金项目:国家自然科学基金(81601461),湖北省自然科学基金(2017CFB552)
作者单位E-mail
谌先敢 中南民族大学 生物医学工程学院 chenxg@mail.scuec.edu.cn 
刘海华 中南民族大学 生物医学工程学院  
李亮 武汉大学人民医院放射科
武汉大学人民医院放射科 
 
潘宁 中南民族大学 生物医学工程学院  
摘要点击次数: 113
全文下载次数: 0
中文摘要:
      在前列腺癌的诊断过程中,候选病灶的检测是一项重要步骤,该步骤有时由医生手工完成,这会带来一些问题.为了实现候选病灶的自动检测,我们训练了一个分类模型用于自动检测候选病灶.获取候选病灶之后,病灶区域的各类特征被用来表征候选病灶,其中,纹理特征在诊断过程中已经被证实是有效的,为了进一步提升性能,候选病灶的更高水平的特征仍然被需要.因此,我们设计了新特征来描述候选病灶:病灶-凸包比,为了证实该特征的有效性,我们设计了实验:在加入新特征之前和之后分别测试计算机辅助检测方法的性能.实验结果表明,我们设计的新特征有助于提升该方法的性能.
英文摘要:
      In the diagnosis of prostate cancer, the detection of candidate lesions is an important step, which is sometimes delineated by experienced radiologist manually. This may bring observer variability. In order to achieve automatic detection of candidate lesions, we train a classification model for automatic detection of candidate lesions. After obtaining the candidate lesion, various features of lesions are used to characterize the candidate lesion, in which the texture feature has been proven to be effective in the diagnosis process. In order to further enhance the performance, the high-level description of tumor candidate lesions is still need. Therefore, we designed a new feature to describe the candidate lesions: the ratio of lesion and convex hull. In order to confirm the effectiveness of the new feature, we designed the experiment: before and after adding new features, the performance of the system was tested. The experimental results show that the new features can improve the performance of the system.
View Fulltext   查看/发表评论  下载PDF阅读器
关闭