Heart Murmur Detection using Machine Learning

MSc-in-AI-Demokritos-Machine-Learning-Course

---

Authors

Name	Registration number
Boura Tatiana	MTN2210
Sideras Andreas	MTN2214

Contents of this repository:

• Report.pdf - The report of our project.

• Presentation.pdf - The presentation of our project.

• the-circor-digiscope-phonocardiogram-dataset-1.0.3 - The dataset.

• The_CirCor_DigiScope_Dataset.pdf - The dataset's paper.

• requirements.txt - Requirements.

• notebooks - Folder that includes:

  The Jupyter Notebooks,

  • Feature_Extraction_&_Demonstration.ipynb, where feature extraction and demonstration is performed.
  • Feature_Selection.ipynb, where the train, validation, test sets are created and feature selection is performed.
  • murmur_classification.ipynb, where we select the ML model, train and evaluate it.

  The Python modules:

  • data_loader_ML.py, that is used from Feature_Extraction_&_Demonstration.ipynb to load the dataset.
  • feature_extraction_ML.py, that is used from Feature_Extraction_&_Demonstration.ipynb to extract the audio features.

  The dataset containing the features extracted from Feature_Extraction_&_Demonstration.ipynb:

  • murmor_dataset.csv

  A script demo_murmur.py that runs a server on localhost for demonstration purposes. Run demo_murmur.py without any arguments after installing the packages in requirements.txt.

  The following two pickle files that store the selected model and the standard scaler respectively. Both files are created from murmur_classification.ipynb.

  • final_model.pkl
  • scaler.pkl

• train_val_test_datasets - Folder that includes the the train, validation, test sets are stored from Feature_Selection.ipynb.

• important_features - Folder that includes .txt files where each feature selection method stores its most important features Feature_Selection.ipynb.

• classifiers_results - Folder that includes .txt files where murmur_classification.ipynb stores for each feature selection, the selected models' classifiers results are stored.

Process

In order to run the whole process you should execute the notebooks,

1. Feature_Extraction_&_Demonstration.ipynb
2. Feature_Selection.ipynb
3. murmur_classification.ipynb

with the given order.

However, every notebook can also be executed separately.

Otherwise, you could just run the demo demo_murmur.py (check requirements.txt), where you can manually select .wav audios for the four regions (AV,MV,PV,TV) from the folder training_data and test our final model.