CF患者肺组织分类——基于相关反馈的监督核哈希方法

浏览次数： 11

分享到:

摘要:

现有的哈希方法用于CF患者肺组织分类时没有从正负反馈样本挖掘判别信息,分类精度不高.为此,提出一种基于相关反馈的监督核哈希方法.首先,对肺组织进行监督核哈希学习,得到初始哈希函数;其次,使用该初始哈希函数对肺组织进行哈希编码和分类,并得到正负反馈样本;接着,基于正负反馈样本构建新的哈希函数;最后,使用新构建的哈希函数对肺组织再次进行哈希编码和分类.实验结果表明,同现有方法相比,所提出的方法显著提高了CF患者肺组织的分类精度.

Performance of traditional hashing methods applied to classification of lung tissues from patients with Cystic Fibrosis (CF) disease is not satisfied, because these methods do not explore discriminant information contained in positive and negative samples obtained from classification results. In this paper, we propose a new kernel-based supervised hashing method boosted by relevance feedback. Firstly, a preliminary hashing function is learned by performing kernel-based supervised hashing on lung tissues. Secondly, the learned hashing function is utilized to encode and classify lung tissues. Then positive and negative samples are obtained from the classification results. Thirdly, a new version of hashing function is learned from these feedback examples. Finally, the newly learned hashing function is employed to classify lung tissues again. Experimental results show that, compared with other existing hashing methods, the proposed method greatly enhances classification results of lung tissue from patients with CF.

作者:

申华磊邱鹏

Shen Hualei;Qiu Peng(College of Computer & Information Engineering,Henan Normal University,Xinxiang 453007,China)

机构地区:

betway官方app 计算机与信息工程学院

出处:

《betway官方app 学报：自然科学版》 CAS 北大核心 2019年第4期24-30,共7页

基金:

国家自然科学基金(61502319) 河南省科技攻关项目(172102210337 182102210363) betway官方app 博士启动经费支持课题(qd16120)

关键词:

相关反馈监督核哈希肺组织分类 CT影像

relevance feedback kernel-based supervised hashing lung tissue classification CT image