This project uses python 2.7

The aim of this project is to classify heartbeats into four categories: normal, murmur, extra heart sound and artifact.

Two datasources will be used to solve this classification problem.

The first comes from the 2016 PhysioNet/CinC Challenge. The data can be found here .

The second source comes from Kaggle. The data can be found here. This dataset is too small so the first dataset will be used to train a more generalize model. This model will only classify between normal and abnormal heartbeats. From the knowledge learned from this model a more specialized model will be trained to classify between normal, murmur, extra heart sound and artifact heartbeats.

The following folder structure has to be followed to have the same results and for the methods in the notebooks to work properly.

data/

physionet/
    reduced_features/
    training-a/
    training-b/
    ...
    training-f/

kaggle/
    reduced_features/
    set_a.csv
    set_a/

---

This project will consist of several notebooks

WB1-explore.ipynb

• This notebook contains code to visualize audio files
• Explores the Physionet and Kaggle datasets

WB2-FNN.ipynb

• This notebook implements a Feed Forward Network for the PhysioNet dataset

WB3-CNN.ipynb

• This notebook implements a Convolutional Neural Network for the PhysioNet dataset

WB4-Transer Learning.ipynb

• This notebook implements the methods applied for feature extraction in WB2 and WB3 and compares the results to model that uses transfer learning to be used with the Kaggle dataset