ECG Deep Learning Hands-on — Building Biosignal AI Models with the MIT-BIH Arrhythmia Dataset
A practical, ethics-first path from WFDB signal processing to 1D CNN classification and patient-level evaluation.
Foundation track (M0–M2)
Research Ethics and Licensing
Distinguish open from credentialed data, preserve attribution, and follow a practical credentialing checklist.
Reading and Visualizing Signals with WFDB
Understand the WFDB format and safely download, inspect, and visualize MIT-BIH records.
ECG Preprocessing
Understand baseline wander, band-pass filtering, resampling, and common evaluation traps.
Getting Started with Colab
From the Google Colab interface to your first ECG plot, with slides and English narration.
Opening soon
Deep-dive track (M3–M8)
R-peak Detection (Pan–Tompkins)
Build the R-peak detection pipeline and validate it against reference annotations.
AAMI Five-class Classification (1D CNN)
Design a 1D CNN experiment that accounts for severe class imbalance.
The Inter-patient Split Trap
Reproduce the misleading performance caused by patient overlap.
RR-interval Rhythm Classification
Design an RR-interval rhythm-classification workflow with AFDB.
Signal Quality Assessment
Check signal quality and missingness before analysis.
Reproducibility and Experiment Management
Record data versions, splits, parameters, and results for reproducible experiments.