TDA_EXTRACTOR

Heart Sound Classification with DTHF Features and CNN

This project is focused on classifying heart sound recordings using features extracted through DTHF (Duration Time Homology Features) and a CNN (Convolutional Neural Network) model.

Datasets

Please download the 2016 PhysioNet/Computing in Cardiology Challenge dataset from the following link:

https://archive.physionet.org/challenge/2016/

Once downloaded, extract the dataset and place the files in the dataset directory within this project. The directory structure should look like this:

project_root/
│
├── dataset/
│   ├── training-a/
│   ├── training-b/
│   ├── training-c/
│   ├── training-d/
│   ├── training-e/
│   ├── training-f/
│   ├── validation/
│   └── REFERENCE.csv
├── extract_DTHF.py
├── main.py
└── ...

Feature Extraction

To extract DTHF features from the dataset, run the extract_DTHF.py script. This will process the heart sound recordings and generate feature files for training and validation:

python extract_DTHF.py

This script will generate .npy files containing the extracted features, which will be saved in the same directory as the dataset.

Training the Model

After the features have been extracted, you can train the CNN model using the extracted DTHF features. Simply run the main.py script:

python main.py

This will train a CNN model for classifying heart sound recordings as either normal or abnormal, and the trained model will be saved in the project directory.