Case Study: Aalto University achieves first place in PhysioNet/CinC Challenge for atrial fibrillation detection with MathWorks' MATLAB

MATLAB Based Algorithm Wins the 2017 PhysioNet/CinC Challenge to Automatically Detect Atrial Fibrillation

Aalto University researchers (part of the Black Swan team) entered the 2017 PhysioNet/Computing in Cardiology Challenge to develop an algorithm that detects atrial fibrillation (AF) and other rhythms from noisy, short single‑lead ECG recordings. They used MathWorks' MATLAB and associated toolboxes (Signal Processing, Wavelet, Statistics and Machine Learning, Parallel Computing, Deep Learning) to analyze signals, extract features, and build classifiers to address limitations of handheld ECG data.

Using MathWorks' MATLAB platform, the team visualized ECGs, applied wavelet and signal‑processing filters, and extracted nearly 500 features (later reduced to 150 via feature‑ranking) with parallelized computation on a 48‑core workstation. They trained and evaluated multiple models and selected a random forest classifier that achieved 81.9% (10‑fold CV) on training data and 82.6% on unseen test data, earning a first‑place tie in the challenge—demonstrating MathWorks tools sped development, visualization, and parallel feature extraction.

Aalto University

Ali Bahrami Rad

Postdoctoral Researcher