P53immunostained cell nuclei segmentation in tissue images of oral squamous cell carcinoma
详细信息 查看全文
- 作者：K. A. Shahul Hameed ; A. Banumathi ; G. Ulaganathan
- 关键词：Gray ; level co ; occurrence matrix ; Global entropy ; Thresholding ; Overlapped/touching cell nuclei ; Watershed transform
- 刊名：Signal, Image and Video Processing
- 出版年：2017
- 出版时间：February 2017
- 年：2017
- 卷：11
- 期：2
- 页码：363-370
- 全文大小：
- 刊物类别：Engineering
- 刊物主题：Signal,Image and Speech Processing; Image Processing and Computer Vision; Computer Imaging, Vision, Pattern Recognition and Graphics; Multimedia Information Systems;
- 出版者：Springer London
- ISSN：1863-1711
- 卷排序：11
文摘
This paper presents segmentation of p53immunostained tissue images of oral squamous cell carcinoma that consist of cell nuclei segmentation and splitting of overlapping/touching cell structures. In segmentation, the entropy thresholding has been adopted in which the optimum threshold value to each color component of the image is obtained by maximizing the global entropy computed from its gray-level co-occurrence matrix. The segmented image consists of isolated cells and complex nuclei structures. A novel complex nuclei structure detection algorithm is proposed to identify overlapped nuclei structures, which have been further resolved by watershed transform. The performance of the segmentation technique is evaluated using the quantitative metrics, namely mean absolute difference (MAD), dice coefficient (DC) and accuracy. Global entropy thresholding-based segmentation technique achieved the best MAD of 0.478, DC of 0.967 and accuracy of 0.970 compared to state-of-art techniques such as otsu and active contour. Extensive experimental results show that proposed complex nuclei structure detection-based overlapping/touching cells splitting algorithm effectively delineated nuclei with over- and under-segmentation rate of 0.49 %. Therefore, tissue image segmentation method presented has high potential in immunohistochemical (IHC) quantification and also can be easily generalized for images stained with other biomarkers.