Publication: Easing scoring in ER and Ki-67 breast cancer histopathological images
Loading...
Identifiers
Publication date
2012-01-10
Defense date
2012-07-10
Authors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
A technique for easing breast cancer scoring in inmunohistochemically stained tissue images is proposed. The method is based on the statistical information extracted from manual scores performed on a collection of images. The main purpose of the thesis is to base the cell counting on nuclei size statistics and using a series of sampling masks avoiding processing the entire biopsy, which normally is the most time consuming part when analysing this kind of images. In order to achieve these results, a dictionary is learnt using a training image. After this step is completed, the dictionary is applied on the test image so its segmentation is obtained. The different elements that compose the image are then differentiated and labelled attaining to three different classifications: blue nuclei, brown nuclei or background area. Finally, scoring is performed following the clinical methods for the ER and Ki–67 staining biomarkers. Relationship between segmented pixels and nuclei size is used in order to estimate the score. The technique conclusions try to demonstrate that close results can be achieved to the ones obtained with the manual counting scoring method employed by pathologists. Thesis concludes with the implementation of a scoring graphical user interface which gathers all the information, algorithms and methods used along the research. This tool tries to provide future collaborators with a really close and helpful instrument eluding to make them go through endless code lines. This way, since day one they will be able to obtain scores and start thinking about the goals they want to accomplish.
Description
Keywords
Proceso de imágenes, Biopsias, Cancer, Histología