A Novel High-Throughput Multispectral Cell Segmentation Algorithm
An increasingly common component of molecular diagnostics is the analysis of protein localization at the single-cell level using fluorescent microscopy. Manually extracting quantitative data from a large population of cells is unreasonably time-consuming and existing automatic systems are rather limited.
Here we present an integrated image analysis software system for high-throughput segmentation of cells and corresponding nuclei. The system is composed of robust image enhancement, followed by multispectral identification of putative cells using a statistical model (Gaussian Mixture Model) approach, followed by cross-spectral watershed that effectively segments clustered cells, and finally, a rule based refinement using statistical morphological attributes of the cells.
The robustness and accuracy of the system have been tested on artificial fluorescent beads, as well as on hand segmented and visually inspected images. Lastly, we compare our algorithm to state-of-the-art systems and show it does better on most performance parameters.
To date, the system has been used to accumulate data from over 300 million segmented cells each expressing specific set of genomic alterations.
KeywordsCancer Functional genomics Live-cells Gaussian mixture model Watershed Cross-spectral image segmentation
The authors would like to gratefully acknowledge the entire NovellusDx team which run the biological experiments and the deep discussions with regards to biological parameters to exclude cells. The authors would also like to thank Prof Zohar Yakhini for his careful review and insightful comments.
- 5.Coelho, L.P., Shariff, A., Murphy, R.F.: Nuclear segmentation in microscope cell images: a hand-segmented dataset and comparison of algorithms. In: Proceedings/IEEE International Symposium on Biomedical Imaging: from Nano to Macro IEEE International Symposium on Biomedical Imaging, vol. E6, pp. 518–521 (2009). doi: 10.1109/ISBI.2009.5193098
- 9.Otsu, N.: A threshold selection method from gray level histogram? IEEE Trans. Syst. Man Cybern. SMC–8, 62–66 (1978)Google Scholar
- 13.Lim, J.S.: Two-Dimensional Signal and Image Processing. Prentice Hall, Englewood Cliffs (1990). Equations 9.26, 9.27, and 9.29Google Scholar
- 14.Poynton, C.: Digital Video and HDTV Algorithms and Interfaces. Morgan Kaufman Publishers, San Francisco (2003)Google Scholar