Development of a Stained Cell Nuclei Counting System
Oral Defence Date:
Professors Kaz Okada, Ilmi Yoon, and Chris Moffatt (Biology)
This thesis presents a novel cell nuclei counting system which exploits a machine vision algorithm called Fast Radial Symmetry Transformation (FRST). FRST is designed to detect points of interest such as eyes in machine vision problems and we used this algorithm to enhance the occurrences of stained-cell nuclei in 2D digital images. The nuclei are detected and then counted using image processing algorithms, such as image binarization, mathematical morphological operations and connected-component analysis. Failure of existing cell counting applications to perform accurately with the images that have non-uniform foreground and background is the motivation for our application. Our system significantly raises the accuracy of the cell counts in the images that have non-uniform foreground and background. The data available and quantitative experiments conducted so far indicates that introduction of FRST algorithm improves the accuracy of cell counting and our system, which has FRST in its core, is at least ten times better than the algorithms without FRST.
Automated cell counting, Cell quantification, Image processing, Fast radial symmetry transformation