nickarad / atrial-classification

Deep Neural Network for Atrial Fibrillation Classification using Tensorflow (For education purpose only)

Install WFDB Software Package

 Follow instructions in: https://physionet.org/physiotools/wfdb-linux-quick-start.shtml 

Install requirements

> pip install -r requirements.txt
> apt-get install python-tk

Install requirements

> pip install -r requirements.txt

Download Dataset

> wget -r -np http://www.physionet.org/physiobank/database/afpdb/

Copy paste afpdb folder to current directory

> mv www.physionet.org/physiobank/database/afpdb .

Go to src folder

> cd src

Create CSV files

> python dat2csv.py

Create training and test dataset

> mkdir training test

Create .npy files

>  python create_dataset.py

Create Neural Network model and train it

> python train.py

Best Model

 Rename the best model from my_model.h5 to best.h5

Make predictions with best model (90% accuracy)

> python predict.py

Tensoarboard

> tensorboard --logdir=logs/

Accuracy & Loss

Normal Record

Abnormal Record

atrial-classification's Issues

Better project structure?

Really good stuff, but the root directory seems bloated.

Maybe move the PNGs to an images directory and the code to a src directory (or similar names? Cheers!