The following python code does analysis of histopathological cancer whole slide images which are in a very high resolution .svs format. It involves image manipulation and extraction of nuclear and cellular for automated detection of cancer
Details:
1.) Our input images are the histopathological images downloaded from TCGA database. These images are known as whole slide images and are in .svs format. They are very high resolution images and therefore their size is significantly high and so these .svs images cannot be processed or viwed using the default image viwer present in our system.
2.)So the first step in our image analysis protocol is to split the large .svs image into many smaller low resolution .png images known as tiles.This makes our image processing easier, as now instead on working on the whole .svs image (which would have been tedious) we would be doing our processing on each of the tiles that we have generated from our parent image.
3.)After succesfully splitting the .svs images we then need to perfrom certain fundamental image manipulating methods which are color normalisation, color deconvolution, nuclei segmentation.
4.)Color Normalisation: There might be some distortions in the image due to uneven lighting and normalisation helps us to get rid of such distortions as well as shadows. Our input for this step in the splitted png tile.
5.)Color Deconvolution:The input for this step is the normalised png tile. In this step we convert the RGB channels of our image into Haemotoxylin(nuclear stain) and Eosin(cytoplasmic stain) channels because these two were the stains that were used to stain the tissue that is bieng shown in the image.
6.)Nuclei Segmentation:The input for this step is the color deconvoluted png image. In this step we identify and isolate the nuclei that are present in our image. It is the most crucial step for efiicient feature extraction.
7.) After the nuclei segmentation step our next and the final step is the extraction of certain features from these png images and perform statistical operations on them.
8.)We extracted ten features for each cell in the split image. The seven types of morphological and spatila features of cell nuclei were: major axis length(Major_Axis), minor axis length(Minor_Axis), the ratio of major to minor axis length(Ratio), nuclear area(Area), mean distance to neighbouring cells(Mean_Distance), maximum distance to neughbouring cells(Max_Distance), minimum distance to neughbouring cells(Min_distance) and the other three features were red, blue and green pixel trait at the centroid of each nuclei of each cell.
9.)After successfull extraction of all these ten features we performed certain statistical operations on each these features. The statistical operations were average, variance, standard deviation, skew, kurtosis and entropy. So in total we have ten features and each of the ten features have five statistical operations associated with them.
###The code also generates a GUI in which the user just have to add the path of the input folder which contains all the sub folders containing .svs whole slide images of cancer.
###Important: Please refer to this installation manual before using the code -
Histomicstk Introduction and Installation: Histomicstk is an open source software(package) which contains algorithms for fundamental image analysis tasks such as color normalization, color deconvolution, cell-nuclei segmentation, and feature extraction on our histopathological svs images. This software can be installed in two ways:
1.) As a pure python package: When we want to use histomicstk within python. 2.) As a server side Girder plug in for web based analysis: When we want to use histomicstk over the web.
Since we have used python throughout this project therefore we installed histomicstk as a pure python package.
Before instaliing histomicstk for linux system the following requirements should be checked for smooth installtion of histomistk:
1.)It is recommended to have a python version 3(64-bit) to avoid unnecessary errors while installation.
2.)We installed the following libraries using the commands below as they will be required during our image analysis:
$ sudo pip3 install matplotlib
$ sudo pip3 install scikit-image
$ sudo pip3 install numpy
$ sudo pip3 install large_image
3.)Histomicstk uses the large_image library for reading and manipulating large svs images but for certain formats such as our svs format large_image also requires an additional package(dependency) which is the OpenSlide library.
4.)So for successfully installing OpenSlide in our system, following commands were used:
$ sudo apt-get install openslide-tools
$ sudo apt-get install python-openslide
$ sudo pip3 install openslide-python
5.)Now, finally we installed Histomicstk using pip:
$ sudo pip3 install histomicstk
Another way of installing Histomicstk is directly from source using the following commands:
$ git clone https://github.com/DigitalSlideArchive HistomicsTK.git
$ cd HistomicsTK
$ sudo pip3 install -e .
6.)We can check whether the packages have been successfully installes by trying to import them by using command import module name