ChestX is a TechLabs Berlin project that aims to detect COVID-19 infection by processing chest X-ray images. To support medical decision making, the ChestX team wants to apply AI techniques for object detection & classification, as well as data science methods. Metadata provided alongside the X-ray images can be analysed to provide doctors with more meaningful diagnostic assistance.
The COVID-19 image data collection is a public open dataset of chest X-ray and CT images of patients which are positive or suspected of COVID-19 or other viral and bacterial pneumonias. The project was started and is maintained by Joseph Paul Cohen, Paul Morrison, and Lan Dao from the University of Montreal.
Please, watch the following video for a quick project introduction, or check out their paper for more details about the data set.
To complement the data we found suitable after data-cleaning, x-ray images from this covid19 radiography database were also added to this project for AI training.
The COVID-19 data sets is linked as submodule to this git project.
Run the following code to download the submodule:
git submodule update --init --recursive
To update the submodul run:
git submodule update --remote --merge
Do download the second data set, please run this CML to connect to the kaggle API:
kaggle datasets download -d tawsifurrahman/covid19-radiography-database
Referencing the COVID-19 image data collection repository, some ML models were developed alongside the data collection process and constitute the main purpose of this research project. A valuble source is the repo from Linda Wang and her team. The repo contains pre-trained models.
Other noteworthy projects using this data set:
- https://github.com/aildnont/covid-cxr
- Covid-19-covid CNN model - Github repo
- Covid-scanner Web-App
- Colab Notebook
Noteworthy papers & research:
Ongoing Kaggle competition based on another data set:
- https://www.kaggle.com/c/siim-covid19-detection/data The images are available in a medical image format - DICOM or /*.dcm
After clonning the repository, and having the complete project files saved in a directory on you computer, open the project folder with your desired editor.
To run the frontend part of the project:
- Navigate to the directory of the frontend files:
cd frontend-chest-X
- Install node.js on your computer if you do not already have it installed node.js
- To check the version of node.js you have installed, type:
node -v
- Install angular globally:
npm install -g @angular/cli
- To install all the frontend dependencies:
npm install
- To open the frontend on localhost:
ng serve --port 4200
Setting up the database
- Install MongoDB on windows
- Install MongoDB on mac
- Install MongoDB compass for managing the database
To run the backend part of the project:
From the project folder st21-chestX
- To install all the backend dependencies:
npm install
- run
npm i nodemon
- To run the backend server:
nodemon server.js
ornode server.js
- DS | Heike
- DS | Jason Choufani
- DS | Assumpta Ojukwu
- DS | FabianB
- AI | Kevin G.
- AI | Ehiz Ali
- UX | Sofia Sabarini
- UX | Nina Thom
- WD | Omobolaji Koyi
- WD | Martin Schöne
- Aleksandr Shubenkov
- Python3
- Python libs: pandas, numpy, matplotlib, (tf.image, tf.keras), pillow, etc.
- Fastai
- Git & GitHub
- Colab notebooks
- PyCharm (or other IDE - personal choice)
- venv
- Angular
- Node.js and Express
- MongoDB and Mongoose
- WIP ...