/RPNet

A Deep Learning approach for robust R Peak detection in noisy ECG.

Primary LanguageJupyter NotebookGNU General Public License v3.0GPL-3.0

RPNet

A Deep Learning approach for robust R Peak detection in noisy ECG

Research

Architecture Diagram

Datasets

  1. Chinese Physiological Signal Challenge (CPSC 2019)
  2. MITBIH Database
  3. MIT-BIH ST Change Database
  4. Noise Stress Test DataBase (NSTDB)

Quantitative Comparisons

Evaluation of model and three traditional methods on CPSC dataset

Evalation of Model on the other 3 datasets

Evaluation of Model in presence of noise(SNR wise) on NSTDB

Qualitative Results

Steps

-- General Steps

  • Download all the 4 datasets.

-- To train the model

  • Run train_CPSC.ipynb

-- To Evaluate the model

  • Download the model
  • To evaluate on CPSC: sh evaluate_detectors_CPSC.sh
  • To evaluate on any one of the other three datasets: sh evaluate_detectors_CPSC.sh
  • To evaluate on the NSTDB dataset: sh evaluate_nstdb.sh

Details on possible changes that can be made to the scipt will be mentioned in the script. We would like to acknowledge Mr Bernd Porr whose repo we forked for the implementation of the traditional ECG Detectors.