Research of a Cervical Cell Sorter Based on BP Neural Network

Early cervical cancer screening is an effective method for reducing the morbidity and mortality of cervical cancer. But the traditional method of artificial cervical cancer screening has the disadvantages of high false negative and false positive. Hence a novel method is proposed in this paper to realize the automatic classification of cervical pathologic cells, while a computer aided diagnosis system is designed. A BP neural network is applied to design a cervical cell sorter that is the key component of the computer-aided cervical cancer screening system. The cervical cell sorter is implemented in Visual C++ 6.0 platform to determine the classification parameters based on the experiments, and to achieve neural network learning based on a momentum-adaptive learning rate adjustment algorithm which is proposed by combining momentum method with an adaptive algorithm of the learning rate. By extracting 15 morphological characters of cervical cells, four types of cervical pathologic cells can be recognized well.