homogenPy

homogenPy is a python pipeline to perform homogeneity assessment of MSI dataset(s). Our pipeline contains different functions to understand drug distribution by taking advantage from different texture analysis based methods, such as an intensity histogram based first-order statistics, gray-level co-occurence matrix (GLCM) based, gray-level run length matrix (GLRLM) based, size-zone matrix (SZM) based and shape factors based.

The pipeline relies mainly on numpy, scipy, skimage, argparse python modules. There are some specific dependencies as well, discussed below. Pipeline main methods are :

1) GetIonImage : import and visualization of data from Analyze 7.5 format. This function offers ion intensity map at user defined mass range (.sim), three different types of segmentation map : a) drug mask(default, _drug.msk, drug mask.jpg) based on either drug mass range or tissue mass range b) tic image mask (_tic.msk, _ticmsk.jpg) c) maximum intensity value mask(_mim.msk, _mimmsk.jpg)

2) GCLM_features : This function will return 13 Haralick texture features based on gray-level co-occurence matrix for input image. Before features calculation, tissue image will multiple with corresponding mask image to place tissue object on uniform background. Hence results were obtained from tissue object only.

3) GLRLM_features : This function will return 11 features based on gray-level run-length matrix.

4) SZM_features : This function will return 11features calculated on size-zone based matrix. Required : rpy2 python module and r-library radiomics

5) SB_features : This function will calculate shape based features for tissue mask image. It will return: number of small disconnected objects within tissue, their area and perimeter. Required : cv2 python module.

{

How to use it :

All scripts run as command line argument. Since output files obtained from GetIonIntensity method will use as an input for other methods, hence it is important to run this script first. After that feature calculation can be done in any order.

To get help about method, use -h argument. It will give details about input arguments and other optional arguments.

Let's start with GetIonInensity

This method creates an ion intensity image and its mask image at desired m-z value. In our study, we had knowledge about m-z value associate with drug compound, tissue object and internal standard respectively. Hence, multiple m-z values can be passed as an input argument. In the absence of any argument, ion intensity image at complete m-z scale will be constructed.

mridula@mridula-HP-ProBook-6460b:~/data/$ python GetIonImage.py -h 

usage: GetIonImage.py [-h] [-mz MASSRANGE [MASSRANGE ...]]
                           [-mz_tissue MASSRANGE_TISSUE [MASSRANGE_TISSUE ...]]
                           [-mz_std MASSRANGE_STD [MASSRANGE_STD ...]]
                           [-fr MFILTRAD] [-tic] [-mim]

Create an extracted ion image and its corresponding binary mask

optional arguments:

 -h, --help    show this help message and exit

-mz MASSRANGE [MASSRANGE ...]
             Desired m/z range

-mz_tissue MASSRANGE_TISSUE [MASSRANGE_TISSUE ...]
              m/z range correspond to tissue

-mz_std MASSRANGE_STD [MASSRANGE_STD ...]
              m/z range for standard

-fr MFILTRAD  Radius of the median filter

-tic          TIC based tissue identification

-mim          Maximum Intensity Ion tissue identification

Run examples:

a. Create ion intensity image at desired mass range

           python GetIonImage.py -mz 284.15 284.17

b. Create ion intensity image at desired mass range with standard

          ~/data/$ python GetIonImage.py -mz 284.15 284.17 -mz_std 289.16 289.18

c. Create ion intensity image at desired mass range for drug compound, tissue object and standard, respectively

         ~/data/$ python GetIonImage.py -mz 284.15 284.17 -mz_tissue 281.31 281.33 -mz_std 289.16 289.18

output image files:

Features calculation:

a) First order statistics

    python FOS_features.py -n 8

b) Gray-level co-occurrence matrix based features: 

         python GCLM_features.py -d 5 -n 16

c) Size-sone matrix based features

        SZM_features.py -n 8

d) Gray-level run length matrix based features

         GLRLM_features.py -n 8

e) Shape based features

      python SB_features.py