OMNI / Git / Diff of /README.md

Models:

tobiasharvey/

OMNI

Downloads: 1

Diff of /README.md [c0487b] .. [5b5f7c]

Switch to unified view


# OMNI (Open Source Monitoring of Neonates and Infants) 







## Software Requirements
Run `sh requirements.sh` in a virtual environment in order to download the required libraries. 

## Hardware Requirements (Edge implementation)

![Edge HW block diag](https://user-images.githubusercontent.com/1295467/76793911-5c573680-679c-11ea-8eb9-fe0abab7e5ac.png)
### Components needed:

 1. [OpenBCI Ganglion kit](https://shop.openbci.com/products/ganglion-board?variant=13461804483)
 
     Open source biosignal acquisition hardware for [research grade biosignal acquisition](https://openbci.com/community/published-research-with-openbci/)
 
 2. [Raspberry Pi 4](https://www.raspberrypi.org/products/raspberry-pi-4-model-b/)
 3. [Rasberry Pi 4 cooling case](https://www.newegg.com/p/1W8-00Y1-00032)
 4. Shirt using [conductive textile electrodes](https://www.alibaba.com/product-detail/Conductive-textile-ecg-Electrodes_1127697682.html) [1$ for 1 electrode]
 

Wearable ECG electrodes --> OpenBCI Ganglion ---(Bluetooth)---> Raspberry Pi 4

## System Configuration
* Ubuntu 16.04
* Nvidia 1080Ti - (Required for training the model)

## Train a model to extract R peaks and Heart Rate from ECG waveform.  
* Download ECG MITDB monitoring data from https://storage.googleapis.com/mitdb-1.0.0.physionet.org/mit-bih-arrhythmia-database-1.0.0.zip and unzip it.
* To train: `python train.py --preprocess_data --data_path "PATH TO DATA"` 

## Train a model to extract Breathing Rate from ECG waveform. 
* Download ECG from the BIDMC database which is derived from MIMIC-II from https://physionet.org/static/published-projects/bidmc/bidmc-ppg-and-respiration-dataset-1.0.0.zip and unzip it. 
* To train: `python train.py --preprocess_data --data_path "PATH TO DATA"`

## Model Inference
* Download ECG from the preterm infant database from https://physionet.org/static/published-projects/picsdb/preterm-infant-cardio-respiratory-signals-database-1.0.0.zip and unzip it. 
* Download the Heart Rate computation model from here: https://drive.google.com/open?id=1yI7G4nofjuzFWkD1CfsOtLZxaukTu0di
* Download the Breathing Rate computation model from here: https://drive.google.com/open?id=1ycV74LfGmgcGmLlrPn2VileeFNsGrRZT
* To run inference and view GUI type: `python run_model.py --path_dir "PATH TO DATA" --saved_hr_model_path "PATH TO HR MODEL" --saved_br_model_path "PATH TO BR MODEL" --patient_no 3 --viewer 1`

# OMNI OpenBCI Pi Inference

 Edge inference of ECG R-peak detection and Respiration extraction using Raspberry Pi 4 using ECG (OpenBCI Ganglion).


## Hardware Design

Wearable ECG electrodes --> OpenBCI Ganglion ---(Bluetooth)---> Raspberry Pi 4

## Software Design

OpenBCI client ----(LSL)--->  Python -> PyTorch inference --> Breathing Rate, Heart Rate


## Installation instruction

### Install PyTorch on Raspberry Pi 4:

 1. Install PyTorch dependicies 
 `sudo apt install libopenblas-dev libblas-dev m4 cmake cython python3-yaml libatlas-base-dev`
 2. Increase swap file memory to 1600, Edit variable `CONF_SWAPSIZE` in `/etc/dphys-swapfile`
 3. Reset environmental variables like ONNX_ML [Instructions](https://gist.github.com/akaanirban/621e63237e63bb169126b537d7a1d979) 
 4. Download PyTorch package compiled for Armv7 ([torch-1.1.0-cp37-cp37m-linux_armv7l.whl](https://github.com/marcusvlc/pytorch-on-rpi/blob/master/torch-1.1.0-cp37-cp37m-linux_armv7l.whl))
 5. Install using the command `sudo pip3 install torch-1.1.0-cp37-cp37m-linux_armv7l.whl` in the same directory

Refer [here](https://github.com/marcusvlc/pytorch-on-rpi) for troubleshooting 

### Install OpenBCI Ganglion client on Raspberry Pi 4:

1. Clone OpenBCI_Python repo
 `git clone htps://github.com/OpenBCI/OpenBCI_Python.git`
2. Install the following requisites python packages using `pip3 install`
    pylsl, python-osc, six, socketIO-client, websocket-client, Yapsy, xmldict, bluepy
3. Open folder OpenBCI_Python and run   
    `sudo python3 user.py --board ganglion -a streamer_lsl` to open a lab streaming layer stream of sensor data from the ganglion
    
## Instructions to run to perform real time breathing rate/ heart rate inference using OpenBCI data
1. Run the lsl streamer script to get data to the inference script
`sudo python3 user.py --board ganglion -a streamer_lsl`
3. Run the visualization and edge inference code on the pi using  `python3 lsl_openbci.py`


## Sample Predictions
![image](https://user-images.githubusercontent.com/1295467/76792925-742dbb00-679a-11ea-8865-6d5f8c83cfe6.png?raw=true)


Above is an example prediction for noisy real time ECG data obtained using the edge inference model. The beat predictions are represented as blue markers on the ECG. 

	a/README.md		b/README.md
1	# OMNI (Open Source Monitoring of Neonates and Infants)	1	# OMNI (Open Source Monitoring of Neonates and Infants)
2		2
3	<p align="center">	3
4	<image src = 'images/omni-logo.png' >	4
5	</p>	5
6		6
7
8
9	## Software Requirements	7	## Software Requirements
10	Run `sh requirements.sh` in a virtual environment in order to download the required libraries.	8	Run `sh requirements.sh` in a virtual environment in order to download the required libraries.
11		9
12	## Hardware Requirements (Edge implementation)	10	## Hardware Requirements (Edge implementation)
13		11
14	![Edge HW block diag](https://user-images.githubusercontent.com/1295467/76793911-5c573680-679c-11ea-8eb9-fe0abab7e5ac.png)	12	![Edge HW block diag](https://user-images.githubusercontent.com/1295467/76793911-5c573680-679c-11ea-8eb9-fe0abab7e5ac.png)
15	### Components needed:	13	### Components needed:
16		14
17	1. [OpenBCI Ganglion kit](https://shop.openbci.com/products/ganglion-board?variant=13461804483)	15	1. [OpenBCI Ganglion kit](https://shop.openbci.com/products/ganglion-board?variant=13461804483)
18		16
19	Open source biosignal acquisition hardware for [research grade biosignal acquisition](https://openbci.com/community/published-research-with-openbci/)	17	Open source biosignal acquisition hardware for [research grade biosignal acquisition](https://openbci.com/community/published-research-with-openbci/)
20		18
21	2. [Raspberry Pi 4](https://www.raspberrypi.org/products/raspberry-pi-4-model-b/)	19	2. [Raspberry Pi 4](https://www.raspberrypi.org/products/raspberry-pi-4-model-b/)
22	3. [Rasberry Pi 4 cooling case](https://www.newegg.com/p/1W8-00Y1-00032)	20	3. [Rasberry Pi 4 cooling case](https://www.newegg.com/p/1W8-00Y1-00032)
23	4. Shirt using [conductive textile electrodes](https://www.alibaba.com/product-detail/Conductive-textile-ecg-Electrodes_1127697682.html) [1$ for 1 electrode]	21	4. Shirt using [conductive textile electrodes](https://www.alibaba.com/product-detail/Conductive-textile-ecg-Electrodes_1127697682.html) [1$ for 1 electrode]
24		22
25		23
26	Wearable ECG electrodes --> OpenBCI Ganglion ---(Bluetooth)---> Raspberry Pi 4	24	Wearable ECG electrodes --> OpenBCI Ganglion ---(Bluetooth)---> Raspberry Pi 4
27		25
28	## System Configuration	26	## System Configuration
29	* Ubuntu 16.04	27	* Ubuntu 16.04
30	* Nvidia 1080Ti - (Required for training the model)	28	* Nvidia 1080Ti - (Required for training the model)
31		29
32	## Train a model to extract R peaks and Heart Rate from ECG waveform.	30	## Train a model to extract R peaks and Heart Rate from ECG waveform.
33	* Download ECG MITDB monitoring data from https://storage.googleapis.com/mitdb-1.0.0.physionet.org/mit-bih-arrhythmia-database-1.0.0.zip and unzip it.	31	* Download ECG MITDB monitoring data from https://storage.googleapis.com/mitdb-1.0.0.physionet.org/mit-bih-arrhythmia-database-1.0.0.zip and unzip it.
34	* To train: `python train.py --preprocess_data --data_path "PATH TO DATA"`	32	* To train: `python train.py --preprocess_data --data_path "PATH TO DATA"`
35		33
36	## Train a model to extract Breathing Rate from ECG waveform.	34	## Train a model to extract Breathing Rate from ECG waveform.
37	* Download ECG from the BIDMC database which is derived from MIMIC-II from https://physionet.org/static/published-projects/bidmc/bidmc-ppg-and-respiration-dataset-1.0.0.zip and unzip it.	35	* Download ECG from the BIDMC database which is derived from MIMIC-II from https://physionet.org/static/published-projects/bidmc/bidmc-ppg-and-respiration-dataset-1.0.0.zip and unzip it.
38	* To train: `python train.py --preprocess_data --data_path "PATH TO DATA"`	36	* To train: `python train.py --preprocess_data --data_path "PATH TO DATA"`
39		37
40	## Model Inference	38	## Model Inference
41	* Download ECG from the preterm infant database from https://physionet.org/static/published-projects/picsdb/preterm-infant-cardio-respiratory-signals-database-1.0.0.zip and unzip it.	39	* Download ECG from the preterm infant database from https://physionet.org/static/published-projects/picsdb/preterm-infant-cardio-respiratory-signals-database-1.0.0.zip and unzip it.
42	* Download the Heart Rate computation model from here: https://drive.google.com/open?id=1yI7G4nofjuzFWkD1CfsOtLZxaukTu0di	40	* Download the Heart Rate computation model from here: https://drive.google.com/open?id=1yI7G4nofjuzFWkD1CfsOtLZxaukTu0di
43	* Download the Breathing Rate computation model from here: https://drive.google.com/open?id=1ycV74LfGmgcGmLlrPn2VileeFNsGrRZT	41	* Download the Breathing Rate computation model from here: https://drive.google.com/open?id=1ycV74LfGmgcGmLlrPn2VileeFNsGrRZT
44	* To run inference and view GUI type: `python run_model.py --path_dir "PATH TO DATA" --saved_hr_model_path "PATH TO HR MODEL" --saved_br_model_path "PATH TO BR MODEL" --patient_no 3 --viewer 1`	42	* To run inference and view GUI type: `python run_model.py --path_dir "PATH TO DATA" --saved_hr_model_path "PATH TO HR MODEL" --saved_br_model_path "PATH TO BR MODEL" --patient_no 3 --viewer 1`
45		43
46	# OMNI OpenBCI Pi Inference	44	# OMNI OpenBCI Pi Inference
47		45
48	Edge inference of ECG R-peak detection and Respiration extraction using Raspberry Pi 4 using ECG (OpenBCI Ganglion).	46	Edge inference of ECG R-peak detection and Respiration extraction using Raspberry Pi 4 using ECG (OpenBCI Ganglion).
49		47
50		48
51	## Hardware Design	49	## Hardware Design
52		50
53	Wearable ECG electrodes --> OpenBCI Ganglion ---(Bluetooth)---> Raspberry Pi 4	51	Wearable ECG electrodes --> OpenBCI Ganglion ---(Bluetooth)---> Raspberry Pi 4
54		52
55	## Software Design	53	## Software Design
56		54
57	OpenBCI client ----(LSL)---> Python -> PyTorch inference --> Breathing Rate, Heart Rate	55	OpenBCI client ----(LSL)---> Python -> PyTorch inference --> Breathing Rate, Heart Rate
58		56
59		57
60	## Installation instruction	58	## Installation instruction
61		59
62	### Install PyTorch on Raspberry Pi 4:	60	### Install PyTorch on Raspberry Pi 4:
63		61
64	1. Install PyTorch dependicies	62	1. Install PyTorch dependicies
65	`sudo apt install libopenblas-dev libblas-dev m4 cmake cython python3-yaml libatlas-base-dev`	63	`sudo apt install libopenblas-dev libblas-dev m4 cmake cython python3-yaml libatlas-base-dev`
66	2. Increase swap file memory to 1600, Edit variable `CONF_SWAPSIZE` in `/etc/dphys-swapfile`	64	2. Increase swap file memory to 1600, Edit variable `CONF_SWAPSIZE` in `/etc/dphys-swapfile`
67	3. Reset environmental variables like ONNX_ML [Instructions](https://gist.github.com/akaanirban/621e63237e63bb169126b537d7a1d979)	65	3. Reset environmental variables like ONNX_ML [Instructions](https://gist.github.com/akaanirban/621e63237e63bb169126b537d7a1d979)
68	4. Download PyTorch package compiled for Armv7 ([torch-1.1.0-cp37-cp37m-linux_armv7l.whl](https://github.com/marcusvlc/pytorch-on-rpi/blob/master/torch-1.1.0-cp37-cp37m-linux_armv7l.whl))	66	4. Download PyTorch package compiled for Armv7 ([torch-1.1.0-cp37-cp37m-linux_armv7l.whl](https://github.com/marcusvlc/pytorch-on-rpi/blob/master/torch-1.1.0-cp37-cp37m-linux_armv7l.whl))
69	5. Install using the command `sudo pip3 install torch-1.1.0-cp37-cp37m-linux_armv7l.whl` in the same directory	67	5. Install using the command `sudo pip3 install torch-1.1.0-cp37-cp37m-linux_armv7l.whl` in the same directory
70		68
71	Refer [here](https://github.com/marcusvlc/pytorch-on-rpi) for troubleshooting	69	Refer [here](https://github.com/marcusvlc/pytorch-on-rpi) for troubleshooting
72		70
73	### Install OpenBCI Ganglion client on Raspberry Pi 4:	71	### Install OpenBCI Ganglion client on Raspberry Pi 4:
74		72
75	1. Clone OpenBCI_Python repo	73	1. Clone OpenBCI_Python repo
76	`git clone htps://github.com/OpenBCI/OpenBCI_Python.git`	74	`git clone htps://github.com/OpenBCI/OpenBCI_Python.git`
77	2. Install the following requisites python packages using `pip3 install`	75	2. Install the following requisites python packages using `pip3 install`
78	pylsl, python-osc, six, socketIO-client, websocket-client, Yapsy, xmldict, bluepy	76	pylsl, python-osc, six, socketIO-client, websocket-client, Yapsy, xmldict, bluepy
79	3. Open folder OpenBCI_Python and run	77	3. Open folder OpenBCI_Python and run
80	`sudo python3 user.py --board ganglion -a streamer_lsl` to open a lab streaming layer stream of sensor data from the ganglion	78	`sudo python3 user.py --board ganglion -a streamer_lsl` to open a lab streaming layer stream of sensor data from the ganglion
81		79
82	## Instructions to run to perform real time breathing rate/ heart rate inference using OpenBCI data	80	## Instructions to run to perform real time breathing rate/ heart rate inference using OpenBCI data
83	1. Run the lsl streamer script to get data to the inference script	81	1. Run the lsl streamer script to get data to the inference script
84	`sudo python3 user.py --board ganglion -a streamer_lsl`	82	`sudo python3 user.py --board ganglion -a streamer_lsl`
85	3. Run the visualization and edge inference code on the pi using `python3 lsl_openbci.py`	83	3. Run the visualization and edge inference code on the pi using `python3 lsl_openbci.py`
86		84
87		85
88	## Sample Predictions	86	## Sample Predictions
89	![image](https://user-images.githubusercontent.com/1295467/76792925-742dbb00-679a-11ea-8865-6d5f8c83cfe6.png)	87	![image](https://user-images.githubusercontent.com/1295467/76792925-742dbb00-679a-11ea-8865-6d5f8c83cfe6.png?raw=true)
90		88
91	![image](images/ezgif.com-gif-maker.gif)
92	Above is an example prediction for noisy real time ECG data obtained using the edge inference model. The beat predictions are represented as blue markers on the ECG.	89	Above is an example prediction for noisy real time ECG data obtained using the edge inference model. The beat predictions are represented as blue markers on the ECG.