Various Techniques for Classification and Segmentation of Cervical Cell Images - A Review

Abstract

Pap smear test plays an important role for the early diagnosis of cervical cancer in which human cells taken from the cervix of patient are analysed for pre-cancerous changes. The manual analysis of these cells by expert cytologist is labor intensive and time consuming job. The automatic and accurate detection of cervical cells are two critical preprocessing steps for automatic Pap smear image analysis and also for diagnosis of pre-cancerous changes in the uterine cervix. Similarly, the reliable segmentation of abnormal nuclei in cervical cytology is of utmost importance in automation-assisted screening techniques. This paper presents, the existing automated methods for the detection, segmentation and boundary determination of cells nuclei in conventional Pap stained cervical smear images. The majority of cytoplasm segmentation uses K-means algorithm, edge detection method, thresholding approach, graph cut and active contours technique. Most of existing work is done on images of isolated cells, especially for those which are in the Herlev data set. For segmentation of images which contains multiple cells, level set and thresholding techniques have been used. The nucleus segmentation varies as: single-nucleus segmentation, touching-nuclei splitting and multiple-nuclei segmentation. However, many segmentation methods incorporates shape priors, usually enforcing elliptical shapes in order to overcome cell occlusion and noise. The main focus of this paper is comprehensive literature survey of various existing classification and segmentation techniques. The shortcomings and failures of the existing work are also provided for further enhancement and improvement of overall performance and accuracy.

Ant Colony Optimization-Enabled CNN Deep Learning Technique for Accurate Detection of Cervical Cancer

“…The patch-based sorting allowed for the calculation of the central value. According to Sharma and Mangat [13], an additional improvement was made to the work, the author improved the "fuzzy c-means (FCM)" clustering technique by working with various numbers of clusters rather of just one. This was done in order to reinforce the clustering of the data.…”

Automated Detection and Classification of Cervical Cancer Using Pap Smear Microscopic Images: A Comprehensive Review and Future Perspectives

“…[107] B. Sharma et al propose an idea of integrating segmentation methods with the region of interest locating methods to improve the segmentation and classification accuracy and achieve an accuracy of 93.7 % with Herlev dataset [108] D. A. Clausi et al use Gray Level Co-occurrence Probabilities (GLCP) for segmenting images and concluded that the GLCPs are more sensitive to texture boundary which results in good classification accuracy. [109] W William et al present a review on image analysis and machine learning techniques for automated cervical cancer screening and identified K-nearest-neighbors and support vector machines algorithms as an excellent classifiers for cervical image with an accuracies of over 99.27% and 98.5% respectively for 2class classification problem.…”

A novel automation-assisted cervical cancer reading method based on convolutional neural network

“…While automation-assisted reading systems can increase productivity by reducing the time needed to read slides, their current performance and costs are not recommended for application in primary cervical screening [11,12]. To this end, lots of automation-assisted methods based on cervical cell image analysis have been proposed [5,13,14,15,16].…”