Segmentation and abnormality detection of cervical cancer cells using fast elm with particle swarm optimization

Abstract

Cervical cancer arises when the anomalous cells on the cervix mature unmanageable obviously in the renovation sector. The most probably used methods to detect abnormal cervical cells are the routine and there is no difference between the abnormal and normal nuclei. So that the abnormal nuclei found are brown in color while normal nuclei are blue in color. The spread or cells are examined and the image denoising is performed based on the Iterative Decision Based Algorithm. Image Segmentation is the method of paneling a digital image into compound sections. The major utilize of segmentation is to abridge or modify the demonstration of an image. The images are segmented by applying anisotropic diffusion on the Denoised image. Image can be enhanced using dark stretching to increase the quality of the image. It separates the cells into all nuclei region and abnormal nuclei region. The abnormal nuclei regions are further classified into touching and non-touching regions and touching regions undergoes feature selection process. The existing Support Vector Machines (SVM) is classified few nuclei regions but the time to taken for execution is high. The abnormality detected from the image is calculated as 45% from the total abnormal nuclei. Thus the proposed method of Fast Particle Swarm Optimization with Extreme Learning Machines (Fast PSO-ELM) to classify all nuclei regions further into touching region and separated region. The iterative method for to training the ELM and make it more efficient than the SVM method. In experimental result, the proposed method of Fast PSO-ELM may shows the accuracy as above 90% and execution time is calculated based on the abnormality (ratio of abnormal nuclei regions to all nuclei regions) image. Therefore, Fast PSO-ELM helps to detect the cervical cancer cells with maximum accuracy.