This repository contains the codes corresponding to study continous blood pressure estimation using PPG features.

Please install the following Python (3.8) libraries:

• hrv-analysis
• biosppy
• pyampd
• scipy
• mat73
• sklearn
• pandas
• numpy
• matplotlib
• pickle

You can skip steps 1,2, and 3 and use the pre-extracted features.csv.

1-

• Download MIMIC II dataset (BP, ECG, PPG) raw signals and save as MATLAB matrices in "data" folder. (similar to the matrices available on UCI BP dataset)

or,

• Use the sample dataset provided in data folder and load it using the following code:

>>> import pickle
>>> with open('data_sample_BP_PPG.pkl', 'rb') as f:
...   [data] = pickle.load(f)

2- Run featureExtractor.py

3- Run CSVconcat.py to concat CSV feature files created by the previous step.

4- Run BPEstimation.py to fit and test machine learning based models.

The results of BP estimation using Random Forest model (10-fold cross validation, MIMIC II dataset):

Systolic BP: corr=0.79, MAE=11.12 mmHg, STD=9.96 mmHg

Diastolic BP: corr=0.78, MAE=4.06 mmHg, STD=3.95 mmHg

Please use the following citation:

Hasanzadeh, Navid, Mohammad Mahdi Ahmadi, and Hoda Mohammadzade. "Blood pressure estimation using photoplethysmogram signal and its morphological features." IEEE Sensors Journal 20, no. 8 (2019): 4300-4310.