基于特征融合和判别式学习的胎盘成熟度自动分级

发布时间：2018-03-26 19:16

本文选题：胎盘成熟度评估　切入点：特征融合　出处：《中国生物医学工程学报》2016年04期

【摘要】：胎盘成熟度分级错误可能会导致小于胎龄儿、死产、死胎等的发生。目前,胎盘成熟度分级主要依赖于临床医生的经验和观察,主观性很强,分级准确性易受医生工作强度、工作时长和工作经验的影响。提出一种基于特征融合和判别式学习的胎盘成熟度自动分级算法。首先,对共544例的B型超声图像和彩色能量多普勒(CDE)胎盘图像,采用提取关键点、对关键点提取特征、进行融合并加以判别式特征编码的方法,形成码书,经过归一化,最后用支持向量机(SVM)进行分类,得到胎盘成熟度分级结果。在测试阶段,将胎盘成熟度测试结果与临床医生的分级结果进行对比,得到如下结果:准确率92.7%,敏感性91.1%,特异性97.6%,平均精度97.3%。结果表明,该方法对胎盘成熟度自动分级具有较高的指导意义。
[Abstract]:At present, placental maturity classification mainly depends on the clinician's experience and observation, and is very subjective, and the accuracy of classification is easy to be subjected to the work intensity of the doctor, so the placental maturity classification may lead to the occurrence of smaller than gestational age, stillbirth, stillbirth, and stillbirth. An automatic grading algorithm for placental maturity based on feature fusion and discriminant learning was proposed. First, 544 cases of B-mode ultrasound images and color power Doppler CDE placental images were studied. Using the method of extracting the key points, extracting the features of the key points, fusion and discriminating feature coding, the codebook is formed. After normalization, the classification is done by support vector machine (SVM). The results of placental maturity grading were obtained. In the testing stage, the results of placental maturity test were compared with those of clinicians, and the following results were obtained: accuracy 92.7, sensitivity 91.1, specificity 97.6. This method has high guiding significance for automatic grading of placental maturity.
【作者单位】：深圳大学医学部生物医学工程学院广东省生物医学信息检测和超声成像重点实验室;深圳妇幼保健院超声科南方医科大学附属医院;
【基金】：深圳市重点基础研究项目(JCYJ2014041415513200;JCYJ20140509172609164) 广东省科技创新重点项目(2014KXM052)
【分类号】：R445.1;R714.56

【相似文献】