Semi-automated pseudo colour features extraction technique for cervical cancer's pap smear images

Image analysis is one of the common applications in the medical field especially in cytology, where the microscopic examination of cells and tissues is involved. Visual interpretation of microscopic images is tedious and in many cases is errorprone. Therefore a number of attempts have been carried out using the computer vision system to supplement the human visual inspection and to automate some of these tedious visual screening tasks. This study, in effect, proposes a semi-automated method of identifying features for Pap smear cytology images; i.e. semi-automated Pseudo-Colour Feature Extraction (PCFE) technique by integrating a clustering algorithm with the manual PCFE algorithm. The technique is used to segment the cervical cell images to provide the clearly seen nucleus and cytoplasm regions and then to extract the four features of cervical cells namely the size of nucleus and cytoplasm of cervical cells, as well as their gray level. A correlation test is applied between the data extracted using the proposed algorithm and data extracted manually by cytotechnologists. The technique operates well on cervical cells images with correlation values approaching 1.0, which indicates a strong positive correlation. The analysis also favours the AFKM clustering algorithm as the best clustering algorithm to be used with the PCFE by possessing the strongest relationship in terms of the correlation value. Furthermore, this study proves that the proposed algorithm is suitable and capable to be used to detect and extract features of cervical cells even for the overlapping cervical cells' images.