Automated segmentation algorithm for cervical cell images by employing cuckoo search based ICM

Abstract

Cervical cancer is one of the most serious cancers, as this disease does not show notable symptoms at early stage. Periodical check ups alone can detect this kind of cancer in an early stage. The earlier the detection of disease, the easier it is to prolong the lifespan of the patient. Taking these statements into account and to serve the society, this article proposes a segmentation algorithm for cervical cell images. The objective of this work is attained in two steps such as cervical image pre-processing and segmentation. The cervical image pre-processing intends to remove the noise by employing adaptive median filter. The irregular staining effect of the cervical images is addressed by Haar wavelets. The background subtraction is achieved by bit slice plane technique. The pre-processed images are segmented by means of Intersecting Cortical Model (ICM) and cuckoo search algorithm. The parameters of ICM are fixed by the cuckoo search algorithm, in order to make the algorithm completely automated. The results of this algorithm are satisfactory in terms of accuracy, sensitivity, specificity and precision rates. The performance of the proposed approach is compared with the related state-of-the-art works.