Advanced

Color separation and segmentation of HxE-stained microscopy pathology images

Lergo Perdiguero, Marta LU (2015) In Bachelor's theses in Mathematical Sciences FMAL01 20151
Mathematics (Faculty of Technology) and Numerical Analysis
Abstract
In this study some microscopic medical images are treated. They are stained with H&E, so they mainly have two colors: pink and purple. The main target is to separate these two colors trying three different algorithms (as any of them is perfect). Pixels in the image have to be separated in two different clusters, so each pixel has to be studied for its classification. This fact makes the computation complicated and with a long runtime. After this separation, two images are found as a result.

The next step is executing a segmentation algorithm, so the image pixels can be divided in two groups: background and cells. Depending on the technique a different segmentation algorithm will be executed. Of course, this will provide a binary image... (More)
In this study some microscopic medical images are treated. They are stained with H&E, so they mainly have two colors: pink and purple. The main target is to separate these two colors trying three different algorithms (as any of them is perfect). Pixels in the image have to be separated in two different clusters, so each pixel has to be studied for its classification. This fact makes the computation complicated and with a long runtime. After this separation, two images are found as a result.

The next step is executing a segmentation algorithm, so the image pixels can be divided in two groups: background and cells. Depending on the technique a different segmentation algorithm will be executed. Of course, this will provide a binary image as a result.

Finally, the pros and cons are executed for each algorithm and they are also compared. Not all of them are successful and there is not a best one. (Less)
Popular Abstract
Over the years, the human has tried to create a machine that can simulate
our behavior in every single way. Also to simulate the human vision, but this is definitely not an easy task.
Just think about how complex and various the world around us is and how easily you understand it. Think of how vivid the three dimensional perception is when you look through your window. You can without any roblem distinguish colors, shapes, textures and even subtle variations in translucency and shading. What is more, visually segmenting every single tree would be instant.

That is it. The human visual system is able to recognize images, process them in the brain and finally get information or conclusions out of it: what is background and what is... (More)
Over the years, the human has tried to create a machine that can simulate
our behavior in every single way. Also to simulate the human vision, but this is definitely not an easy task.
Just think about how complex and various the world around us is and how easily you understand it. Think of how vivid the three dimensional perception is when you look through your window. You can without any roblem distinguish colors, shapes, textures and even subtle variations in translucency and shading. What is more, visually segmenting every single tree would be instant.

That is it. The human visual system is able to recognize images, process them in the brain and finally get information or conclusions out of it: what is background and what is foreground, what is water and what is sky, what is shadow and what is color variation. The question now is...But how is this developed with a machine?

Perceptual psychologists have been trying to understand for decades how the human visual system works and actually a complete solution to this puzzle remains elusive.

That is why, there are many techniques on the table that can help us simulate the human eye. By now, none of them is perfect so we are responsible of choosing the one that adapts better to our target of study. (Less)
Please use this url to cite or link to this publication:
author
Lergo Perdiguero, Marta LU
supervisor
organization
course
FMAL01 20151
year
type
M2 - Bachelor Degree
subject
keywords
Image Analysis, segmentation
publication/series
Bachelor's theses in Mathematical Sciences
report number
LUTFMA-4004-2015
ISSN
1654-6229
other publication id
2015:K5
language
English
id
7449631
date added to LUP
2015-09-09 10:58:47
date last changed
2015-09-09 10:58:47
@misc{7449631,
  abstract     = {In this study some microscopic medical images are treated. They are stained with H&E, so they mainly have two colors: pink and purple. The main target is to separate these two colors trying three different algorithms (as any of them is perfect). Pixels in the image have to be separated in two different clusters, so each pixel has to be studied for its classification. This fact makes the computation complicated and with a long runtime. After this separation, two images are found as a result.

The next step is executing a segmentation algorithm, so the image pixels can be divided in two groups: background and cells. Depending on the technique a different segmentation algorithm will be executed. Of course, this will provide a binary image as a result.

Finally, the pros and cons are executed for each algorithm and they are also compared. Not all of them are successful and there is not a best one.},
  author       = {Lergo Perdiguero, Marta},
  issn         = {1654-6229},
  keyword      = {Image Analysis,segmentation},
  language     = {eng},
  note         = {Student Paper},
  series       = {Bachelor's theses in Mathematical Sciences},
  title        = {Color separation and segmentation of HxE-stained microscopy pathology images},
  year         = {2015},
}