Digital Pathological Image Analysis
Aided by Convolutional Neural Network
Predicts Prognosis of Lung Adenocarcinoma Patients

ConvPath Software - User Manual


Dependencies

  • Matlab:
    1.  Install Matlab. (tested environment: R2015a and R2016b)
    2.  Export Matlab path. The path location is based on where you installed Matlab:
      export PATH=$PATH: /Applications/MATLAB_R2015a.app/bin/;
  • R:
    1.  Install Rscript utility.
    2.  Install R with the following packages: kimisc, magic, abind, grid, tiff, xlsx .
  • Data:
    1.  Create a file folder
    2.  Put your image (.svs file) and region of interest (ROI) definition (.xml file) into the folder.
      ps. Each pair of the svs file and the xml file should share the same filename. All of your results will be in that folder.


Runtime Options

Usage:

perl main.pl [Options]

[Options]:
  • -dd:  Set Data Directory, where you put your image files *Required*
  • -roi:  Set Region of Interest
  • -partition:  Set size of deeplearning bunch of image patches; (recommanded 1000)
  • -version:  Report Version information
  • -help:  Display help message
  • -citation:  Display citation format


File Folder Attachment

  • -cell_border:   Images of predicted cell type and region border for sampling region
    * Tumor cells: green
    * Lymphocytes: blue
    * Stroma cells: red
  • -cell_Info:  Cell center locations and predicted types
  • -deeplearning_results:  Txt files with predicted probabilities of three cell types
  • -extracted_features:  Perimeter and size information for each sampling region
  • -patch_extraction_results:
    • -cellLocationInfo:   Cell center coordinates, one file for each sampling region
    • -ImagePatchInfo:   All image patches (One folder for each sampling region)
    • -MarkedImageInfo:   Original sample regions with detected cell center marked
  • -citation:  Display citation format


Contributor

Tao Wang, Faliu Yi, and Shidan Wang
University of Texas Southwestern Medical Center
(Feb., 1, 2017)


Contact

Email: guanghua.xiao@utsouthwestern.edu


Citation

ConvPath: A software tool for lung adenocarcinoma digital pathological image analysis aided by a convolutional neural network. (2019) EBioMedicine. [Link]