651. Segmenting overlapping cell nuclei in digital histopathology images.
- Author
-
Shu J, Fu H, Qiu G, Kaye P, and Ilyas M
- Subjects
- Biomarkers, Tumor metabolism, Colorectal Neoplasms metabolism, Colorectal Neoplasms pathology, Humans, Tissue Array Analysis methods, Tumor Suppressor Protein p53 metabolism, Cell Nucleus pathology, Image Processing, Computer-Assisted methods, Pathology, Clinical methods
- Abstract
Automatic quantification of cell nuclei in immunostained images is highly desired by pathologists in diagnosis. In this paper, we present a new approach for the segmentation of severely clustered overlapping nuclei. The proposed approach first involves applying a combined global and local threshold method to extract foreground regions. In order to segment clustered overlapping nuclei in the foreground regions, seed markers are obtained by utilizing morphological filtering and intensity based region growing. Seeded watershed is then applied and clustered nuclei are separated. As pixels corresponding to stained cellular cytoplasm can be falsely identified as belonging to nuclei, a post processing step identifying positive nuclei pixels is added to eliminate these false pixels. This new approach has been tested on a set of manually labeled Tissue Microarray (TMA) and Whole Slide Images (WSI) colorectal cancers stained for the biomarker P53. Experimental results show that it outperformed currently available state of the art methods in nuclei segmentation.
- Published
- 2013
- Full Text
- View/download PDF