Automatic clump splitting for cell quantification in microscopical images

Authors:
Gloria Díaz;Fabio Gonzalez;Eduardo Romero
Affiliations:
Bioingenium Research Group, National University of Colombia, Bogotá, Colombia;Bioingenium Research Group, National University of Colombia, Bogotá, Colombia;Bioingenium Research Group, National University of Colombia, Bogotá, Colombia
Venue:
CIARP'07 Proceedings of the Congress on pattern recognition 12th Iberoamerican conference on Progress in pattern recognition, image analysis and applications
Year:
2007

Citing 4
Cited 0

Decomposition of digital clumps into convex parts by contour tracing and labelling

Pattern Recognition Letters
Segmentation of Dense Leukocyte Clusters

MMBIA '01 Proceedings of the IEEE Workshop on Mathematical Methods in Biomedical Image Analysis (MMBIA'01)
Multiscale Connected Operators

Journal of Mathematical Imaging and Vision
A rule-based approach for robust clump splitting

Pattern Recognition

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents an original method for splitting overlapped cells in microscopical images, based on a template matching strategy. First, a single template cell is estimated using an Expectation Maximization algorithm applied to a collection of correctly segmented cells from the original image. Next, a process based on matching the template against the clumped shape and removing the matched area is applied iteratively. A chain code representation is used for establishing best correlation between these two shapes. Maximal correlation point is used as an landmark point for the registration approach, which finds the affine transformation that maximises the intersection area between both shapes. Evaluation was carried out on 18 images in which 52 clumped shapes were present. The number of found cells was compared with the number of cells counted by an expert and results show agreement on a 93% of the cases.