uu.seUppsala University Publications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Automatic grading of breast cancer from whole slide images of Ki67 stained tissue sections
Uppsala University, Disciplinary Domain of Science and Technology, Mathematics and Computer Science, Department of Information Technology, Computerized Image Analysis and Human-Computer Interaction. Uppsala University, Science for Life Laboratory, SciLifeLab. (Quantitative Microscopy)ORCID iD: 0000-0002-6699-4015
Uppsala University, Disciplinary Domain of Science and Technology, Mathematics and Computer Science, Department of Information Technology, Computerized Image Analysis and Human-Computer Interaction. Uppsala University, Science for Life Laboratory, SciLifeLab. (Quantitative Microscopy)
Uppsala University, Disciplinary Domain of Science and Technology, Mathematics and Computer Science, Department of Information Technology, Computerized Image Analysis and Human-Computer Interaction.ORCID iD: 0000-0002-1636-3469
2016 (English)Conference paper, Poster (Other academic)
Abstract [en]

Aim

This work describes a proof-of-principle study within the Exchange of Diagnostic Images in Networks (ExDIN) project, for automatic grading of breast cancer from whole slide images of Ki67 stained tissue sections. The idea was to mimic the manual grading process: “The assessment is carried out on invasive cancer within the area with the highest number of Ki67-positive cancer cell nuclei/area (hot spot), containing at least 200 cells.”

Method

  • Color deconvolution to separate the image into brown and blue channels.

  • Extract the 10 subsampled tiles (size corresponding to ~200 cells) with the highest values for pre-defined texture and color features.

  • Analyze these tiles in full resolution and compute the maximum positivity (defined as area of positive cells in relation to total cell area, rather than number of cells, since that will speed up the computations and avoid introducing errors due to over- or under segmentation of connected objects).

     

     

     

     

     

     

     

     

     

     

     

     

     

Figure 1. Illustration of the procedure. Hot spot candidates are extracted from low resolution tiles. Then the final hot spot is selected among the corresponding full resolution versions.

The results show good correlation to manual estimates and the procedure takes ~4 minutes/slide.

Future improvements

  • Rules and features defined using machine learning based on training samples given by pathologists.

  • User interface where suggested regions can be deselected manually.

Place, publisher, year, edition, pages
2016.
National Category
Medical Image Processing
Research subject
Computerized Image Processing
Identifiers
URN: urn:nbn:se:uu:diva-309606OAI: oai:DiVA.org:uu-309606DiVA: diva2:1052279
Conference
4th Nordic Symposium on Digital Pathology
Projects
ExDIN
Available from: 2016-12-06 Created: 2016-12-06 Last updated: 2016-12-21

Open Access in DiVA

No full text

Search in DiVA

By author/editor
Ranefall, PetterWählby, CarolinaBengtsson, Ewert
By organisation
Computerized Image Analysis and Human-Computer InteractionScience for Life Laboratory, SciLifeLab
Medical Image Processing

Search outside of DiVA

GoogleGoogle Scholar

Total: 274 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf