Key result
Machine learning models incorporating positional features achieve up to ~85% F1 scores for AF detection.
Why the study?
The study aimed to identify the most influential heartbeat interval features for atrial fibrillation detection and evaluate the impact of training datasets on model performance.
Does feature engineering using positional features improve machine learning model detection of atrial fibrillation?
Does feature engineering using positional features improve machine learning model detection of atrial fibrillation?
Feature engineering, particularly the inclusion of positional features, improves the performance of machine learning models in detecting atrial fibrillation from ECG data.
May support ML feature engineering for AF detection; leaves open prospective clinical validation before practice change.
In this paper, we focus on feature engineering to detect Atrial Fibrillation by determining whether the heart rhythm has irregularities without patterns. We experiment with a broad spectrum of features derived from the duration of heartbeat-to-heartbeat intervals in the benchmark electrocardiogram databases MIT-BIH Arrhythmia Database, MIT-BIH Atrial Fibrillation Database, and Long Term AF Database. The experiments included position-based features, fluctuation indices, standard deviations, mean average values, Shannon entropy, and statistical measures, such as compressed time series data length. The research questions are to detect the most influential features that result in the best-performing model and the impact of the training dataset. Our approach is to evaluate the model on a completely different dataset from the one it was trained on. Achieved F1 scores vary between 62,32% and 85.08%. The results prove that positional features increase the model's performance.
No takes yet. Share an insight, caveat, or question.
Tudjarski et al. (2025) studied Atrial Fibrillation. Machine learning models with feature engineering was evaluated on Model performance (F1 score). Machine learning models incorporating positional features achieved F1 scores between 62.32% and 85.08% for detecting atrial fibrillation across independent benchmark ECG datasets.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: