Automated segmentation of cell nuclei in cytology pleural fluid images using OTSU thresholding

The automated segmentation of cell nuclei is critical for diagnosis and classification of cancers in pleural fluid. This task is very essential since the morphology of cell nuclei such as the size, shape and stained color are mainly associated with cells proliferation and malignancy. It remains challenging due to the inconsistent stained color, poor contrast, the variety of cells, and the large amount of cells, cell overlapping and other microscopic imaging artifacts. In this paper, we proposed a simple method that is capable to segment the nuclei of the variety of cells from microscopic cytology pleural fluid images. In the proposed method, the original image is enhanced firstly using median filter. Next, the enhanced image is converted into l*a*b* color space and extract l* and b* component. The cell nuclei are segmented using OTSU thresholding as the binary image. Finally, morphological operations are used to eliminate the undesirable artifacts and reconstruct into color segmented image. The proposed method is tested with 25 Papanicolaou (Pap) stained pleural fluid images. The method is relatively simple and the results are very promising.