Overview
The BIDMC dataset is a dataset of electrocardiogram (ECG), pulse oximetry (photoplethysmogram, PPG) and impedance pneumography respiratory signals acquired from intensive care patients. The dataset is intended to be used for evaluating the performance of respiratory rate algorithms, reflecting their potential performance in a real-world critical care environment. This dataset is a very small subset of the Physionet’s MIMIC II Waveform Database Matched Subset, which contains all MIMIC II Waveform Database records that have been associated with the MIMIC II Clinical Database records. Read more about the MIMIC II Waveform Database here.
Details
The dataset comprises 53 8-minute recordings of ECG, PPG, and impedance pneumography signals (sampling frequency, fs = 125 Hz) acquired from adult patients (aged 19-90+, 32 females). Those in the dataset were randomly selected from a larger cohort of patients who were admitted to medical and surgical intensive care units at the Beth Israel Deaconess Medical Center (BIDMC), Boston, USA. Two sets of annotations (performed manually and independently by two researchers) of individual breaths in the impedance pneumography signal can be used to derive reference respiratory rate values.
Further details of the methods used to extract the data from the MIMIC II Database are provided here.
Example
This figure shows an example of the impedance pneumography, ECG and PPG signals contained in the dataset.Publications
The dataset was used in this publication and the publication below for evaluating the performance of different respiratory rate estimation algorithms.
Pimentel, M.A.F. et al. Towards a Robust Estimation of Respiratory Rate from Pulse Oximeters, IEEE Transactions on Biomedical Engineering, 64(8), pp.1914-1923, 2017.Citation
When using this dataset please cite the publication above, and also the standard citation for PhysioNet:Link to download
The dataset is hosted on PhysioNet, from where it can be downloaded in the following formats:
- WFDB (WaveForm DataBase) format: the standard format used by PhysioNet.
- Matlab format: a single file containing all 53 recordings.
- CSV (Comma-Separated Value) format
Reproducing the dataset
The Matlab script used to extract the dataset from the MIMIC II database is available here.