ecg-mit-bih / Git / Diff of /README.md

Models:

ReneeD/

ecg-mit-bih

Downloads: 1

Diff of /README.md [a378de] .. [d42376]

Switch to unified view


[![license](https://img.shields.io/badge/License-GPL%20v3-blue.svg)](./LICENSE)

## ECG classification using MIT-BIH dataset 

This repo is an implementation of https://www.nature.com/articles/s41591-018-0268-3 and https://arxiv.org/abs/1707.01836

and focus on training using a MIT-BIH dataset. If you want to train using CINC or open irhythm data, see the open source which the authors of the original research paper have coded at https://github.com/awni/ecg

Introduction to MIT-BIH dataset at physionet : https://physionet.org/physiobank/database/mitdb/

### Dependency (Updated, April 1, 2025)

- Python == 3.12.9
- Flask==3.1.0
- gevent==24.11.1
- keras==3.9.1
- numpy==2.1.3
- pip-tools==7.4.1
- scikit-learn==1.6.1
- scipy==1.15.2
- six==1.17.0
- tensorflow==2.19.0
- tensorflow-metal==1.2.0
- tqdm==4.67.1
- Werkzeug==3.1.3
- wfdb==4.2.0




### Data setup and train 

I recommend using a vitual enviroment for Python, so run setup.sh in order to install and to activate it. 
```
$ git clone https://github.com/physhik/ecg-mit-bih.git
$ cd ecg-mit-bih
$ python -m venv ecg-env
$ source ./ecg-env/bin/activate
(ecg-env) $ pip install -r requrirements.txt
(ecg-env) $ python src/data.py --downloading True
(eng-env) $ python src/train.py
```
Now you have a trained model for ECG classification 


### Test

Predict an annotation of [CINC2017 data](https://physionet.org/challenge/2017/) or your own data(csv file)

It randomly chooses one of data, and predict the slices of the signal.

Run predict.py in the virtual environment we have already set up.
```
(ecg-env) $ python src/predict.py --cinc_download True
```
--cinc_download branch is used at first to download the CINC2017 data.

See src/config.py and customize your parameters or learn better way to train and test 


### Jupyter notebook example

In case, you do not have a GPU above a decent performance, you might be able to use Google colab. Follow the [Jupyter notebook](https://github.com/physhik/ecg-mit-bih/blob/master/src/practice/ecg_mit.ipynb).


### Flask web app

The flask web app is based on the keras-flask-deploy [Github repo](https://github.com/mtobeiyf/keras-flask-deploy-webapp). 

#### Run app.py
```
(ecg-env) $ python src/app.py
```



and choose a csv heart heat signal and click predict, and see the result. 



I have put one csv file in static/asset directory. The first value of the column become sample rate of the web app. If you use your own heart beat csv file, insert the sample at the first, too.   

### Using Docker, Buld and run an image for the ECG trained model.(Not yet updated)


After installation of Docker, 

```
$ docker-compose up -d .  
```



### Introduction to ECG 

I presented a bit more about ECG classfications on my personal blog, http://physhik.github.io 

Find the posts from tags or categories easily.  

### Reference to 

The original research papers
https://www.nature.com/articles/s41591-018-0268-3
https://arxiv.org/abs/1707.01836

The open source by authors
https://github.com/awni/ecg

also noticable 
https://github.com/fernandoandreotti/cinc-challenge2017/tree/master/deeplearn-approach

	a/README.md		b/README.md
1	[![license](https://img.shields.io/badge/License-GPL%20v3-blue.svg)](./LICENSE)	1	[![license](https://img.shields.io/badge/License-GPL%20v3-blue.svg)](./LICENSE)
2		2
3	## ECG classification using MIT-BIH dataset	3	## ECG classification using MIT-BIH dataset
4		4
5	This repo is an implementation of https://www.nature.com/articles/s41591-018-0268-3 and https://arxiv.org/abs/1707.01836	5	This repo is an implementation of https://www.nature.com/articles/s41591-018-0268-3 and https://arxiv.org/abs/1707.01836
6		6
7	and focus on training using a MIT-BIH dataset. If you want to train using CINC or open irhythm data, see the open source which the authors of the original research paper have coded at https://github.com/awni/ecg	7	and focus on training using a MIT-BIH dataset. If you want to train using CINC or open irhythm data, see the open source which the authors of the original research paper have coded at https://github.com/awni/ecg
8		8
9	Introduction to MIT-BIH dataset at physionet : https://physionet.org/physiobank/database/mitdb/	9	Introduction to MIT-BIH dataset at physionet : https://physionet.org/physiobank/database/mitdb/
10		10
11	### Dependency (Updated, April 1, 2025)	11	### Dependency (Updated, April 1, 2025)
12		12
13	- Python == 3.12.9	13	- Python == 3.12.9
14	- Flask==3.1.0	14	- Flask==3.1.0
15	- gevent==24.11.1	15	- gevent==24.11.1
16	- keras==3.9.1	16	- keras==3.9.1
17	- numpy==2.1.3	17	- numpy==2.1.3
18	- pip-tools==7.4.1	18	- pip-tools==7.4.1
19	- scikit-learn==1.6.1	19	- scikit-learn==1.6.1
20	- scipy==1.15.2	20	- scipy==1.15.2
21	- six==1.17.0	21	- six==1.17.0
22	- tensorflow==2.19.0	22	- tensorflow==2.19.0
23	- tensorflow-metal==1.2.0	23	- tensorflow-metal==1.2.0
24	- tqdm==4.67.1	24	- tqdm==4.67.1
25	- Werkzeug==3.1.3	25	- Werkzeug==3.1.3
26	- wfdb==4.2.0	26	- wfdb==4.2.0
27		27
28		28
29		29
30		30
31	### Data setup and train	31	### Data setup and train
32		32
33	I recommend using a vitual enviroment for Python, so run setup.sh in order to install and to activate it.	33	I recommend using a vitual enviroment for Python, so run setup.sh in order to install and to activate it.
34	```	34	```
35	$ git clone https://github.com/physhik/ecg-mit-bih.git	35	$ git clone https://github.com/physhik/ecg-mit-bih.git
36	$ cd ecg-mit-bih	36	$ cd ecg-mit-bih
37	$ python -m venv ecg-env	37	$ python -m venv ecg-env
38	$ source ./ecg-env/bin/activate	38	$ source ./ecg-env/bin/activate
39	(ecg-env) $ pip install -r requrirements.txt	39	(ecg-env) $ pip install -r requrirements.txt
40	(ecg-env) $ python src/data.py --downloading True	40	(ecg-env) $ python src/data.py --downloading True
41	(eng-env) $ python src/train.py	41	(eng-env) $ python src/train.py
42	```	42	```
43	Now you have a trained model for ECG classification	43	Now you have a trained model for ECG classification
44		44
45		45
46	### Test	46	### Test
47		47
48	Predict an annotation of [CINC2017 data](https://physionet.org/challenge/2017/) or your own data(csv file)	48	Predict an annotation of [CINC2017 data](https://physionet.org/challenge/2017/) or your own data(csv file)
49		49
50	It randomly chooses one of data, and predict the slices of the signal.	50	It randomly chooses one of data, and predict the slices of the signal.
51		51
52	Run predict.py in the virtual environment we have already set up.	52	Run predict.py in the virtual environment we have already set up.
53	```	53	```
54	(ecg-env) $ python src/predict.py --cinc_download True	54	(ecg-env) $ python src/predict.py --cinc_download True
55	```	55	```
56	--cinc_download branch is used at first to download the CINC2017 data.	56	--cinc_download branch is used at first to download the CINC2017 data.
57		57
58	See src/config.py and customize your parameters or learn better way to train and test	58	See src/config.py and customize your parameters or learn better way to train and test
59		59
60		60
61	### Jupyter notebook example	61	### Jupyter notebook example
62		62
63	In case, you do not have a GPU above a decent performance, you might be able to use Google colab. Follow the [Jupyter notebook](https://github.com/physhik/ecg-mit-bih/blob/master/src/practice/ecg_mit.ipynb).	63	In case, you do not have a GPU above a decent performance, you might be able to use Google colab. Follow the [Jupyter notebook](https://github.com/physhik/ecg-mit-bih/blob/master/src/practice/ecg_mit.ipynb).
64		64
65		65
66	### Flask web app	66	### Flask web app
67		67
68	The flask web app is based on the keras-flask-deploy [Github repo](https://github.com/mtobeiyf/keras-flask-deploy-webapp).	68	The flask web app is based on the keras-flask-deploy [Github repo](https://github.com/mtobeiyf/keras-flask-deploy-webapp).
69		69
70	#### Run app.py	70	#### Run app.py
71	```	71	```
72	(ecg-env) $ python src/app.py	72	(ecg-env) $ python src/app.py
73	```	73	```
74		74
75	![png](src/static/asset/capture1.png)	75
76
77	and choose a csv heart heat signal and click predict, and see the result.	76	and choose a csv heart heat signal and click predict, and see the result.
78		77
79	![png](src/static/asset/capture2.png)	78
80		79
81	I have put one csv file in static/asset directory. The first value of the column become sample rate of the web app. If you use your own heart beat csv file, insert the sample at the first, too.	80	I have put one csv file in static/asset directory. The first value of the column become sample rate of the web app. If you use your own heart beat csv file, insert the sample at the first, too.
82		81
83	### Using Docker, Buld and run an image for the ECG trained model.(Not yet updated)	82	### Using Docker, Buld and run an image for the ECG trained model.(Not yet updated)
84		83
85		84
86	After installation of Docker,	85	After installation of Docker,
87		86
88	```	87	```
89	$ docker-compose up -d .	88	$ docker-compose up -d .
90	```	89	```
91		90
92		91
93		92
94	### Introduction to ECG	93	### Introduction to ECG
95		94
96	I presented a bit more about ECG classfications on my personal blog, http://physhik.github.io	95	I presented a bit more about ECG classfications on my personal blog, http://physhik.github.io
97		96
98	Find the posts from tags or categories easily.	97	Find the posts from tags or categories easily.
99		98
100	### Reference to	99	### Reference to
101		100
102	The original research papers	101	The original research papers
103	https://www.nature.com/articles/s41591-018-0268-3	102	https://www.nature.com/articles/s41591-018-0268-3
104	https://arxiv.org/abs/1707.01836	103	https://arxiv.org/abs/1707.01836
105		104
106	The open source by authors	105	The open source by authors
107	https://github.com/awni/ecg	106	https://github.com/awni/ecg
108		107
109	also noticable	108	also noticable
110	https://github.com/fernandoandreotti/cinc-challenge2017/tree/master/deeplearn-approach	109	https://github.com/fernandoandreotti/cinc-challenge2017/tree/master/deeplearn-approach