Publication detail

Random Forests Pixel-wise Classification for Detection and Segmentation of Cells in the Images from Holographic Microscope

VIČAR, T. KOLÁŘ, R.

Original Title

Random Forests Pixel-wise Classification for Detection and Segmentation of Cells in the Images from Holographic Microscope

Type

conference paper

Language

English

Original Abstract

Microscopic cell image analysis is widely used by biologists for cell behavior and cell morphology study. In dense cell cultures precise single-cell segmentation is challenging task and it is an important step for automatic cell analysis methods. This work introduces a novel method for robust single cell segmentation of images from holographic microscope. The method is based on pixel-wise classification with random forests for both background segmentation a cell detection, where cell detection image is refined with distance transform based detector. Final single cell segmentation combines both detection and background with seeded watershed. Proposed background segmentation part reaches results similar to other algorithms, but cell detection part of the algorithm is innovative and achieves significantly better result than commonly used detector.

Keywords

cell segmentation, cell detection, random forests, distance transform

Authors

VIČAR, T.; KOLÁŘ, R.

Released

28. 8. 2017

Publisher

Vysoké učení technické v Brně, Fakulta elektrotechniky a komunikačních technologií

Location

Brno

ISBN

978-80-214-5526-9

Book

Proceedings of IEEE Student Branch Conference Mikulov 2017

Edition number

první

Pages from

67

Pages to

70

Pages count

4

URL

BibTex

@inproceedings{BUT138793,
  author="Tomáš {Vičar} and Radim {Kolář}",
  title="Random Forests Pixel-wise Classification for Detection and Segmentation of Cells in the Images from Holographic Microscope",
  booktitle="Proceedings of IEEE Student Branch Conference Mikulov 2017",
  year="2017",
  number="první",
  pages="67--70",
  publisher="Vysoké učení technické v Brně, Fakulta elektrotechniky a komunikačních technologií",
  address="Brno",
  isbn="978-80-214-5526-9",
  url="http://www.radio.feec.vutbr.cz/ieee/userfiles/downloads/archive/2017-Mikulov/Proceedings_Mikulov_2017.pdf"
}