Performance Evolution of Various Wavelets in Cervical Lesion Detection

Cervical cancer is one of most common cancers among women in the world caused by human papilloma virus infection. It develops in the tissue of cervix which connects upper body of the uterus to the vagina. The types of cancer are squamous cell carcinoma, adeno carcinoma and adeno squamous carcinoma based on location of cervix where cancer develops. In this paper, an automatic detection of squamous cell carcinoma in cervical images based on Discrete Wavelet Transform (DWT) and K-Nearest Neighbor (KNN) classifier is described. The energy features are extracted from DWT decomposed image of small area of cervical images. Then the features are fed into KNN classifier to classify whether the given area is normal or cancer affected region. The performance of the proposed system is evaluated by using three wavelets namely bi-orthogonal (bior3.7), Daubechies-8(db8) and Symlet (sym8). Experimental results show the performance of db8 with other wavelets that produces 97.22% average accuracy.