introduction

orange-cxr-classifier is a collection of supervised trained machine learning model, trained to classify images of Chest X-Ray (CXR). All model is trained using posterior-anterior x-ray images from open database of COVID-19 cases chest X-ray published here by Joseph Paul Cohen. This trained model is to be use within Orange Data Mining program.

requirements

  1. Orange Data Mining
  2. Image Analytics Addon for Orange Data Mining

usage

predicting

You need:

  1. Images of CXR you want to analyze
  2. Orange Data Mining with Image Analytics Add-on installed
  3. A trained model (you can get it from trained-model folder in this repository)
  4. Orange Workflow file predict-cxr.ows (you can get this file from this repository)

Steps:

  1. Open Orange Workflow file predict-cxr.ows from within Orange Data Mining program.
  2. Follow instruction from step 1 to 8.

training

Why?

COVID-19 image data collection repository published by Joseph Paul Cohen is frequently updated. Training with updated data can improve model accuracy.

You need:

  1. awk or any awk variation, eg, gawk (awk is a domain-specific language designed for text processing)
  2. COVID-19 image data collection repository (can be downloaded here)
  3. unix or unix-like terminal (on Windows, you can use busybox for windows or cygwin)
  4. Orange Data Mining with Image Analytics Add-on installed
  5. Orange Workflow file train-cxr.ows (you can get this file from this repository)
  6. parsemetadata.awk (you can get this file from training-tool folder in this repository)

Steps:

  1. Extract COVID-19 image data collection repository
  2. Copy parsemetadata.awk into COVID-19 image data collection repository main directory
  3. Launch your unix/unix-like terminal and execute parsemetadata.awk with awk using this command: 
    awk -f parsemetadata.awk metadata.csv
  4. This command will sort all image file found in metadata.csv into appropriate folder for training

  1. If everything goes ok, then you are now ready to start the training from within orange data mining.
  2. Launch orange data mining and open train-cxr.ows workflow file.
  3. Continue following instruction from within the workflow file.

license

Copyright (C) 2020 Mohd Kholid Yaacob (http://mrharmonies.blogspot.com)

This source is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.

This code is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

A copy of the GNU General Public License is available on the World Wide Web at http://www.gnu.org/copyleft/gpl.html. You can also obtain it by writing to the Free Software Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, MA 02110-1335, USA.