MCHeart

This repository provides the overall framework for training and evaluating heart murmur detection systems proposed in 'MCHeart: Multi-Channel based Heart Signal Processing Scheme for Heart Noise Detection using Deep Learning'

Data Access

The training data of the George B. Moody PhysioNet Challenge 2022 can be downloaded from PhysioNet [1]. You can also download it directly using this link or the following command:

wget -r -N -c -np https://physionet.org/files/circor-heart-sound/1.0.3/

How do I run these scripts?

You can install the dependencies for these scripts by creating a Docker image (see below) and running

pip install requirements.txt

You can train your model by running

python train_model.py training_data model

where

training_data (input; required) is a folder with the training data files and
model (output; required) is a folder for saving your model.

Requirements

Tensorflow (tested with version 1.3)
Numpy (tested with version 1.13.1)
Scipy (tested with version 0.19.1)
For openmax experiments you will need to clone https://github.com/abhijitbendale/OSDN into the this directory.

Model Training

python3 main.py \
        --batch_size 200 \
        --num_workers 40 \
        --max_epochs 30 \
        --embedding_dim $embedding_dim \
        --save_dir $save_dir \
        --encoder_name $encoder_name \
        --train_csv_path $train_csv_path \
        --learning_rate 0.001 \
        --encoder_name ${encoder_name} \
        --num_classes $num_classes \
        --trial_path $trial_path \
        --loss_name $loss_name \
        --num_blocks $num_blocks \
        --step_size 4 \
        --gamma 0.5 \
        --weight_decay 0.0000001 \
        --input_layer $input_layer \
        --pos_enc_layer_type $pos_enc_layer_type

Citation

If you use this code please cite the paper.

Useful links

References

[1] Heart Murmur Detection from Phonocardiogram Recordings: The George B. Moody PhysioNet Challenge 2022

@inproceedings{reyna2022heart,
  title={Heart murmur detection from phonocardiogram recordings: The george b. moody physionet challenge 2022},
  author={Reyna, Matthew A and Kiarashi, Yashar and Elola, Andoni and Oliveira, Jorge and Renna, Francesco and Gu, Annie and Alday, Erick A Perez and Sadr, Nadi and Sharma, Ashish and Mattos, Sandra and others},
  booktitle={2022 Computing in Cardiology (CinC)},
  volume={498},
  pages={1--4},
  year={2022},
  organization={IEEE}
}

soyul5458/ReLCNN