Introduction

During the Jedha's bootcamp we had the project to classify ECG's Signal by type of noise.
We was 3 on this project :

• Maxence Di Marcantonio
• Henry Dossot
• David Mougey

ECG Data

We used the data from https://physionet.org/content/butqdb/1.0.0/
(Brno University of Technology ECG Quality Database (BUT QDB))

References

Publication 23/11/2019 by Marisa Faraggi, Karim Sayadi : Time series features extraction using Fourier and Wavelet transforms on ECG data [OCTO article]

Starting guide

Presentation / restitution

• see Pdf presentation
• see Jedha Youtube start on 27''

Visualisation with Streamlit

pip install streamlit
pip install requirements.txt 
streamlit ecg-visualization.py

Data exploration

• see wavelet_analysis_ECG.ipynb

Dataset created

Images below

Neuronal Models

• wavelet_analysis_ECG_model.ipynb (Training with one ECG file)

Satisfying results that give us confidence on predicting noisy signals, results can be improved (see our presentation to discover our areas for improvement

• wavelet_analysis_ECG_model_all_patients.ipynb (Training with 3 ECG file)
• wavelet_analysis_ECG_v1.ipynb (signal 1000ms, 1000Hz)
• wavelet_analysis_ECG_v2.ipynb (signal 2000ms, 250Hz)