LCT image recognition for cervical cells based on BP neural network

Globally, cervical cancer is a kind of common malignant tumor second only to breast carcinoma for women. In China, the morbidity has been dramatically rising with a trend that the patients are younger and younger. Each year, about 30 thousand Chinese females died for this disease. Screening, early detection and treatment are very important in reducing the morbidity and mortality. In this paper, we will classify the segmented single cervical exfoliated cell nuclei using BP neural network. By extracting an optimized feature parameter subset of the numerous candidate parameters of the nucleus with the principal component analysis (PCA) method, the highly statistical correlation between feature parameters that may exists is removed, and the runtime efficiency of the computer aided screening system has been greatly improved, which also leads to a more satisfying recognition result.