宫颈细胞图像的特征提取与识别研究

Research on Cervical Cell Image Feature Extraction and Recognition

宫颈涂片的检查是诊断宫颈癌的最有效手段之一,而传统的宫颈细胞识别系统存在很大的局限,例如假阴性率和假阳性率过高。本文为了提高宫颈细胞涂片诊断的效率和准确率,首先提取宫颈细胞的形态特征和极径灰度中值,然后采用AdaBoost-SVM多特征融合分类器进行分类。实验研究结果表明:通过特征提取方法与AdaBoost-SVM多特征融合分类器结合,明显提高了宫颈细胞涂片筛查的效率和准确率,降低了宫颈癌的误诊率。

Cervical smear examination is one of the most effective means of diagnosis of cervical cancer, while the traditional cervical cell recognition system has significant limitations, with low false-negative and false-positive rates. Firstly, morphological characteristics and the gray values of pole in cervical cells are extracted. Then AdaBoost-SVM feature fusion classifier is used to classify the cervical cells in order to improve the efficiency and accuracy of diagnosis of cervical smears. The research results show that the combination of extraction method and multi-feature fusion AdaBoost-SVM classifier can significantly improve the efficiency and accuracy of cervical smear screening, and can reducethe misdiagnosis rate of cervical cancer.