- Introduction
- DISCLAIMER
- Installation
- Train
- Real World Testing
- Citation
- Contributing
- Versioning
- License
- Bugs/Issues
In this research, we have used Matlab and the publicly available SARS-COV-2 Ct-Scan Dataset, containing 1252 CT scans that are positive for SARS-CoV-2 infection (COVID-19) and 1230 CT scans for patients non-infected by SARS-CoV-2. This dataset of CT scans for SARS-CoV-2 (COVID-19) identification is created by our collaborators, Plamenlancaster: Professor Plamen Angelov from Lancaster University/ Centre Director @ Lira, & his researcher, Eduardo Soares PhD.
This project should be used for research purposes only. The purpose of the project is to show the potential of Artificial Intelligence for medical support systems such as diagnosis systems.
Although the program is very accurate and shows good results both on paper and in real world testing, it is not meant to be an alternative to professional medical diagnosis.
Developers that have contributed to this repository have experience in using Artificial Intelligence for detecting certain types of cancer & COVID-19. They are not a doctors, medical, or COVID-19 experts.
Please use this system responsibly.
Please follow the Installation Guide to install COVID-19 xDNN Matlab Classifier.
Assuming you have completed the installation guide, you can now begin training.
Open Matlab 2020 and navigate to the project root directory and execute the following command:
TrainModel
The trained model files can be found in Model/Pretrained location, this contains a MAT file which is trained on 1737 images containing xDNN Parameters. When the training finishes the model will show the results related to metrics and figures of merit.
Elapsed time is 9.566493 seconds.
Results:
Accuracy =
96.6443
precision =
0.9535
recall =
0.9814
F1 =
0.9672
AUC =
0.9663
Confusion_Matrix =
369 7
18 351
Fig 1. Accuracy
Fig 2. ROC/AUC
Fig 3. Precision, Recall and F1
Fig 4. Confusion Matrix
| Accuracy | Recall | Precision | AUC/ROC |
|---|---|---|---|
| 0.966443 | 0.9814 | 0.9535 | 0.9663 |
| Figures of merit | Amount/Value | Percentage |
|---|---|---|
| True Positives | 369 | 49.5302013% |
| False Positives | 7 | 0.9395973% |
| True Negatives | 351 | 47.1140939% |
| False Negatives | 18 | 2.4161073% |
| Misclassification | 25 | 3.3557046% |
| Sensitivity / Recall | 0.9814 | 98% |
The Matlab file DatasetAndFeaturesPreparation.m can be used to make the IMDS and Features MAT files for training on your own dataset. Before running the above script, paste the dataset folder with containing subfolders in the project root directory. After running the above script save the generated imdsTrain, imdsValidation mat files in Imds and featuresTrain, featuresTest in Features Folder.
For testing the model and getting the results on random CT Lung Scan Images, we will upload an Image on a webpage for the given default IP Address and Port in config.json, then it will provide the result for the given Image.The webpage interface to upload and predict images is one using Flask API and we will use python script Server.py, which will call Flask API to interact with the webpage and call the PredictImage.m matlab function through MATLAB Engine API Library which uses the pretrained model to classify the uploaded Image.
To test an Image, navigate to the project root(../xDNN/Projects/Matlab/1/) and execute the following command in command prompt:
python3 Server.py
The Script will start running and initiate both MATLAB Engine API and Flask API.
* Serving Flask app "Server" (lazy loading)
* Environment: production
WARNING: This is a development server. Do not use it in a production deployment.
Use a production WSGI server instead.
* Debug mode: on
* Restarting with stat
* Debugger is active!
* Debugger PIN: 156-368-932
* Running on http://127.0.0.1:5000/ (Press CTRL+C to quit)
Now go to your default browser and search for given HTTP address. You will see a web page as shown below:
Now Upload any JPG, JPEG or PNG CT Scan image file by clicking Upload Image. Keep in Mind that the Image file should have at atleast size of (224,224) image pixels. After Uploading Image, click on Show Results:
The Web Page can take 2-3 minutes to calculate and show the results as the Matlab API engine is run from python. It can take few minutes and performance varies with the given System Configurations.
After Sometime it will give results for the uploaded Image as shown:
Angelov, Plamen, and Eduardo Almeida Soares. "EXPLAINABLE-BY-DESIGN APPROACH FOR COVID-19 CLASSIFICATION VIA CT-SCAN." medRxiv (2020).
Soares, Eduardo, Angelov, Plamen, Biaso, Sarah, Higa Froes, Michele, and Kanda Abe, Daniel. "SARS-CoV-2 CT-scan dataset: A large dataset of real
patients CT scans for SARS-CoV-2 identification." medRxiv (2020). doi: https://doi.org/10.1101/2020.04.24.20078584.
Link:
https://www.medrxiv.org/content/10.1101/2020.04.24.20078584v2
The Peter Moss COVID-19 AI Research Project encourages and welcomes code contributions, bug fixes and enhancements from the Github.
Please read the CONTRIBUTING document for a full guide to forking our repositories and submitting your pull requests. You will also find information about our code of conduct on this page.
-
Aniruddh Sharma - Asociacion De Investigatcion En Inteligencia Artificial Para La Leucemia Peter Moss R&D Junior, Ahmedabad, Gujarat, India
-
Nitin Mane - Asociacion De Investigatcion En Inteligencia Artificial Para La Leucemia Peter Moss AI R&D, Aurangabad, India
We use SemVer for versioning. For the versions available, see Releases.
This project is licensed under the MIT License - see the LICENSE file for details.
We use the repo issues to track bugs and general requests related to using this project. See CONTRIBUTING for more info on how to submit bugs, feature requests and proposals.