Histogram Based Thresholding for Automated Nucleus Segmentation Using Breast Imprint Cytology

Abstract

Breast imprint cytology is a well-recognized technique and provides a magnificent cytological clarity. For imprint cytology slide preparation, tissue samples from the needle taken out to touch and rolled over glass slide and finally stained by hematoxilin and eosin (H&E). The aim of this research is to segment breast imprint cytology nucleus. Images from imprint cytology slides were grabbed by an optical microscope. The histogram based threshold technique has been used to segment nucleus. The proposed technique includes pre-processing, segmentation, post-processing, and final output stage. In pre-processing first image colors was normalized by white balance technique. Then the green channel was extracted from the normalized image. In segmentation stage target nucleus was segmented by pixel intensities. Post-processing stage refers to the clear border nucleus or sharpening the edges. Finally, the three channels were concatenated to get RGB image. The proposed technique performs best in imprint cytology nucleus segmentation, and capable of distinguishing nucleus and non-nucleus objects. Performance of our proposed algorithm is quite high and much more useful for further analysis.