SleepEEGNet / Git / Diff of /README.md

Models:

ReneeD/

SleepEEGNet

Downloads: 1

Diff of /README.md [3f9044] .. [889a15]

Switch to unified view


# SleepEEGNet: Automated Sleep Stage Scoring with Sequence to Sequence Deep Learning Approach
In this study, we introduced a novel deep learning approach, called SleepEEGNet, for automated sleep stage scoring using a single-channel EEG.

# Paper
 Our paper can be downloaded from the [arxiv website](https://arxiv.org/pdf/1903.02108).

 





## Requirements
* Python 2.7
* tensorflow/tensorflow-gpu
* numpy
* scipy
* matplotlib
* scikit-learn
* matplotlib
* imbalanced-learn(0.4.3)
* pandas
* mne
## Dataset and Data Preparation
We evaluated our model using [the Physionet Sleep-EDF datasets](https://physionet.org/physiobank/database/sleep-edfx/) published in 2013 and 2018.  
We have used the source code provided by [github:akaraspt](https://github.com/akaraspt/deepsleepnet) to prepare the dataset.

* To download SC subjects from the Sleep_EDF (2013) dataset, use the below script:

```
cd data_2013
chmod +x download_physionet.sh
./download_physionet.sh
```

* To download SC subjects from the Sleep_EDF (2018) dataset, use the below script:
```
cd data_2018
chmod +x download_physionet.sh
./download_physionet.sh
```

Use below scripts to extract sleep stages from the specific EEG channels of the Sleep_EDF (2013) dataset:

```
python prepare_physionet.py --data_dir data_2013 --output_dir data_2013/eeg_fpz_cz --select_ch 'EEG Fpz-Cz'
python prepare_physionet.py --data_dir data_2013 --output_dir data_2013/eeg_pz_oz --select_ch 'EEG Pz-Oz'
```

## Train

* Modify args settings in seq2seq_sleep_sleep-EDF.py for each dataset.

* For example, run the below script to train SleepEEGNET model with the 20-fold cross-validation using Fpz-Cz channel of the Sleep_EDF (2013) dataset:
```
python seq2seq_sleep_sleep-EDF.py --data_dir data_2013/eeg_fpz_cz --output_dir output_2013 --num_folds 20
```

## Results
* Run the below script to present the achieved results by SleepEEGNet model for Fpz-Cz channel.
```
python summary.py --data_dir output_2013/eeg_fpz_cz
```



## Visualization
* Run the below script to visualize attention maps of a sequence input (EEG epochs) for Fpz-Cz channel.
```
python visualize.py --data_dir output_2013/eeg_fpz_cz
```


## Citation

If you find it useful, please cite our paper as follows:

```
@article{mousavi2019sleepEEGnet,
  title={SleepEEGNet: Automated Sleep Stage Scoring with Sequence to Sequence Deep Learning Approach},
  author={Sajad Mousavi, Fatemeh Afghah and U. Rajendra Acharya},
  journal={arXiv preprint arXiv:1903.02108},
  year={2019}
}
```

## References
 [github:akaraspt](https://github.com/akaraspt/deepsleepnet)  
 [deepschool.io](https://github.com/sachinruk/deepschool.io/blob/master/DL-Keras_Tensorflow)
 
## Licence 
For academtic and non-commercial usage 

	a/README.md		b/README.md
1	# SleepEEGNet: Automated Sleep Stage Scoring with Sequence to Sequence Deep Learning Approach	1	# SleepEEGNet: Automated Sleep Stage Scoring with Sequence to Sequence Deep Learning Approach
2	In this study, we introduced a novel deep learning approach, called SleepEEGNet, for automated sleep stage scoring using a single-channel EEG.	2	In this study, we introduced a novel deep learning approach, called SleepEEGNet, for automated sleep stage scoring using a single-channel EEG.
3		3
4	# Paper	4	# Paper
5	Our paper can be downloaded from the [arxiv website](https://arxiv.org/pdf/1903.02108).	5	Our paper can be downloaded from the [arxiv website](https://arxiv.org/pdf/1903.02108).
6	* The Model architecture	6
7	![Alt text](/images/seq2seq_sleep.jpg)	7
8
9	* The CNN architecture
10
11	![Alt text](/images/seq2seq_cnn.jpg)
12
13	## Requirements	8	## Requirements
14	* Python 2.7	9	* Python 2.7
15	* tensorflow/tensorflow-gpu	10	* tensorflow/tensorflow-gpu
16	* numpy	11	* numpy
17	* scipy	12	* scipy
18	* matplotlib	13	* matplotlib
19	* scikit-learn	14	* scikit-learn
20	* matplotlib	15	* matplotlib
21	* imbalanced-learn(0.4.3)	16	* imbalanced-learn(0.4.3)
22	* pandas	17	* pandas
23	* mne	18	* mne
24	## Dataset and Data Preparation	19	## Dataset and Data Preparation
25	We evaluated our model using [the Physionet Sleep-EDF datasets](https://physionet.org/physiobank/database/sleep-edfx/) published in 2013 and 2018.	20	We evaluated our model using [the Physionet Sleep-EDF datasets](https://physionet.org/physiobank/database/sleep-edfx/) published in 2013 and 2018.
26	We have used the source code provided by [github:akaraspt](https://github.com/akaraspt/deepsleepnet) to prepare the dataset.	21	We have used the source code provided by [github:akaraspt](https://github.com/akaraspt/deepsleepnet) to prepare the dataset.
27		22
28	* To download SC subjects from the Sleep_EDF (2013) dataset, use the below script:	23	* To download SC subjects from the Sleep_EDF (2013) dataset, use the below script:
29		24
30	```	25	```
31	cd data_2013	26	cd data_2013
32	chmod +x download_physionet.sh	27	chmod +x download_physionet.sh
33	./download_physionet.sh	28	./download_physionet.sh
34	```	29	```
35		30
36	* To download SC subjects from the Sleep_EDF (2018) dataset, use the below script:	31	* To download SC subjects from the Sleep_EDF (2018) dataset, use the below script:
37	```	32	```
38	cd data_2018	33	cd data_2018
39	chmod +x download_physionet.sh	34	chmod +x download_physionet.sh
40	./download_physionet.sh	35	./download_physionet.sh
41	```	36	```
42		37
43	Use below scripts to extract sleep stages from the specific EEG channels of the Sleep_EDF (2013) dataset:	38	Use below scripts to extract sleep stages from the specific EEG channels of the Sleep_EDF (2013) dataset:
44		39
45	```	40	```
46	python prepare_physionet.py --data_dir data_2013 --output_dir data_2013/eeg_fpz_cz --select_ch 'EEG Fpz-Cz'	41	python prepare_physionet.py --data_dir data_2013 --output_dir data_2013/eeg_fpz_cz --select_ch 'EEG Fpz-Cz'
47	python prepare_physionet.py --data_dir data_2013 --output_dir data_2013/eeg_pz_oz --select_ch 'EEG Pz-Oz'	42	python prepare_physionet.py --data_dir data_2013 --output_dir data_2013/eeg_pz_oz --select_ch 'EEG Pz-Oz'
48	```	43	```
49		44
50	## Train	45	## Train
51		46
52	* Modify args settings in seq2seq_sleep_sleep-EDF.py for each dataset.	47	* Modify args settings in seq2seq_sleep_sleep-EDF.py for each dataset.
53		48
54	* For example, run the below script to train SleepEEGNET model with the 20-fold cross-validation using Fpz-Cz channel of the Sleep_EDF (2013) dataset:	49	* For example, run the below script to train SleepEEGNET model with the 20-fold cross-validation using Fpz-Cz channel of the Sleep_EDF (2013) dataset:
55	```	50	```
56	python seq2seq_sleep_sleep-EDF.py --data_dir data_2013/eeg_fpz_cz --output_dir output_2013 --num_folds 20	51	python seq2seq_sleep_sleep-EDF.py --data_dir data_2013/eeg_fpz_cz --output_dir output_2013 --num_folds 20
57	```	52	```
58		53
59	## Results	54	## Results
60	* Run the below script to present the achieved results by SleepEEGNet model for Fpz-Cz channel.	55	* Run the below script to present the achieved results by SleepEEGNet model for Fpz-Cz channel.
61	```	56	```
62	python summary.py --data_dir output_2013/eeg_fpz_cz	57	python summary.py --data_dir output_2013/eeg_fpz_cz
63	```	58	```
64		59
65	![Alt text](/images/results.jpg)
66
67	## Visualization	60	## Visualization
68	* Run the below script to visualize attention maps of a sequence input (EEG epochs) for Fpz-Cz channel.	61	* Run the below script to visualize attention maps of a sequence input (EEG epochs) for Fpz-Cz channel.
69	```	62	```
70	python visualize.py --data_dir output_2013/eeg_fpz_cz	63	python visualize.py --data_dir output_2013/eeg_fpz_cz
71	```	64	```
72		65
73		66
74	## Citation	67	## Citation
75		68
76	If you find it useful, please cite our paper as follows:	69	If you find it useful, please cite our paper as follows:
77		70
78	```	71	```
79	@article{mousavi2019sleepEEGnet,	72	@article{mousavi2019sleepEEGnet,
80	title={SleepEEGNet: Automated Sleep Stage Scoring with Sequence to Sequence Deep Learning Approach},	73	title={SleepEEGNet: Automated Sleep Stage Scoring with Sequence to Sequence Deep Learning Approach},
81	author={Sajad Mousavi, Fatemeh Afghah and U. Rajendra Acharya},	74	author={Sajad Mousavi, Fatemeh Afghah and U. Rajendra Acharya},
82	journal={arXiv preprint arXiv:1903.02108},	75	journal={arXiv preprint arXiv:1903.02108},
83	year={2019}	76	year={2019}
84	}	77	}
85	```	78	```
86		79
87	## References	80	## References
88	[github:akaraspt](https://github.com/akaraspt/deepsleepnet)	81	[github:akaraspt](https://github.com/akaraspt/deepsleepnet)
89	[deepschool.io](https://github.com/sachinruk/deepschool.io/blob/master/DL-Keras_Tensorflow)	82	[deepschool.io](https://github.com/sachinruk/deepschool.io/blob/master/DL-Keras_Tensorflow)
90		83
91	## Licence	84	## Licence
92	For academtic and non-commercial usage	85	For academtic and non-commercial usage