End-to-End Myocardial Infarction Classification from Echocardiographic Scans

Published in ASMSUS 2022 [Paper]

This repository contains a PyTorch implementation of our method used in the paper above for MI classification from echocardiography videos using models pretrained on ejection fraction prediction.

Installation

Clone this repository and enter the directory:

git clone https://github.com/BioMedIA-MBZUAI/mi-classification.git
cd mi-classification

The code is implemented for Python 3.8.10.

Install dependencies using pip:

pip install -r requirements.txt

Data

EchoNet-Dynamic (EF Pretraining)

1. Download the dataset from EchoNet-Dynamic website
2. Run the following to extract one cardiac cycle from each video, preprocess and store it as a tensor:

cd ef
python3 echo.py path/to/EchoNet/Videos

CAMUS (EF Pretraining)

1. Download the dataset from CAMUS challenge website
2. Run the following to extract one cardiac cycle from each video, preprocess and store it as a tensor:

cd ef
python3 camus.py path/to/CAMUS/files

HMC-QU (MI Classification)

1. Download the dataset from Kaggle
2. Run the following to extract one cardiac cycle from each video, preprocess and store it as a tensor:

A4C

cd mi/a4c
python3 cycle.py /path/to/A4C/videos/

A2C

cd mi/a2c
python3 cycle.py /path/to/A2C/videos/

Ejection Fraction Pretraining

cd mi/ef
python3 train.py

MI Classification

A4C

cd mi/a4c
python3 train.py --pretrained

A2C

cd mi/a2c
python3 train.py --pretrained

Citation

@inproceedings{saeed2022end,
  title={End-to-End Myocardial Infarction Classification from Echocardiographic Scans},
  author={Saeed, Mohamed and Yaqub, Mohammad},
  booktitle={International Workshop on Advances in Simplifying Medical Ultrasound},
  pages={54--63},
  year={2022},
  organization={Springer}
}