基于超像素和支持向量机的阴道细菌自动检测

发布时间：2018-04-24 14:03

本文选题：阴道细菌 + 超像素　；参考：《中国生物医学工程学报》2015年02期

【摘要】：阴道受到细菌感染引发的阴道炎疾病可能导致异位妊娠、不孕、急慢性盆腔炎等严重疾病,目前形态学人工观察是临床诊断该类疾病的主要方法,但容易引起误诊和漏诊。本研究提出一种基于超像素和支持向量机(SVM)的阴道细菌自动检测方法,对革兰染色的阴道细菌图像,采用简单线性迭代聚类(SLIC)方法计算超像素;对超像素区域计算形状特征、颜色特征和方向梯度直方图(HOG)特征;最后用SVM对超像素区域进行识别。在专业医生的指导下挑选了40幅正常图像和60幅有细菌性阴道病(BV)的图像进行实验,其中10幅正常图像和20幅有细菌性阴道病(BV)的图像用于训练分类器,剩下的70幅用于测试算法。实验结果表明,所提出的自动检测算法获得了89.27%的细菌检出率,具有较大的临床应用价值。
[Abstract]:Vaginitis caused by bacterial infection may lead to ectopic pregnancy, infertility, acute and chronic pelvic inflammatory diseases and other serious diseases. At present, morphological observation is the main method of clinical diagnosis of this disease, but it is easy to cause misdiagnosis and missed diagnosis. In this study, a new method of automatic detection of vagina bacteria based on super-pixel and support vector machine (SVM) is proposed. A simple linear iterative clustering (SLIC-SLIC) method is used to calculate the superpixel and the shape feature of the super-pixel region for gram-stained vaginal bacterial image. Color feature and directional gradient histogram (hog) feature; finally, SVM is used to identify the super-pixel region. Under the guidance of a professional doctor, 40 normal images and 60 BV images with bacterial vaginosis were selected for experiment, of which 10 normal images and 20 BV images with bacterial vaginosis were used to train classifiers. The remaining 70 are used to test the algorithm. The experimental results show that the proposed automatic detection algorithm can obtain 89.27% bacterial detection rate, which has great clinical application value.
【作者单位】：深圳大学医学院生物医学工程系广东省生物医学信息检测与超声成像重点实验室;深圳市第六人民医院;深圳大学附属南山医院;
【分类号】：R711.31;TP181

【参考文献】