Cytoplasm segmentation on cervical cell images using graph cut-based approach

Issue title: Frontiers in Biomedical Engineering and Biotechnology – Proceedings of the 2nd International Conference on Biomedical Engineering and Biotechnology, 11–13 October 2013, Wuhan, China

Article type: Research Article

Authors: Zhang, Ling^{; ;} | Kong, Hui | Chin, Chien Ting^{; ;} | Wang, Tianfu^{; ; ;} | Chen, Siping^{; ; ;}

Affiliations: Department of Biomedical Engineering, Shenzhen University, Shenzhen 518060, China | National-Regional Key Technology Engineering Laboratory for Medical Ultrasound, Shenzhen 518060, China | Guangdong Key Laboratory of Biomedical Information Detection and Ultrasound Imaging, Shenzhen 518060, China | Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139 USA

Note: [] Corresponding author. E-mail: [email protected].

Note: [] Corresponding author. E-mail: [email protected].

Abstract: This paper proposes a method to segment the cytoplasm in cervical cell images using graph cut-based algorithm. First, the A* channel in CIE LAB color space is extracted for contrast enhancement. Then, in order to effectively extract cytoplasm boundaries when image histograms present non-bimodal distribution, Otsu multiple thresholding is performed on the contrast enhanced image to generate initial segments, based on which the segments are refined by the multi-way graph cut method. We use 21 cervical cell images with non-ideal imaging condition to evaluate cytoplasm segmentation performance. The proposed method achieved a 93% accuracy which outperformed state-of-the-art works.

Keywords: Cervical cell, cytoplasm segmentation, A* channel, graph cut

DOI: 10.3233/BME-130912

Journal: Bio-Medical Materials and Engineering, vol. 24, no. 1, pp. 1125-1131, 2014