A Graph-Constrained Changepoint Learning Approach for Automatic QRS-Complex Detection

TL;DR

This paper introduces a graph-based changepoint detection method for automatic QRS-complex detection in ECG signals, utilizing a learned constraint graph to improve accuracy and robustness.

Contribution

It proposes a novel graph learning algorithm that iteratively refines a constraint graph for improved R-peak detection in ECG signals.

Findings

Achieves 99.64% sensitivity in R-peak detection

Attains 99.71% positive predictivity

Reaches a detection error rate of 0.19

Abstract

This study presents a new viewpoint on ECG signal analysis by applying a graph-based changepoint detection model to locate R-peak positions. This model is based on a new graph learning algorithm to learn the constraint graph given the labeled ECG data. The proposed learning algorithm starts with a simple initial graph and iteratively edits the graph so that the final graph has the maximum accuracy in R-peak detection. We evaluate the performance of the algorithm on the MIT-BIH Arrhythmia Database. The evaluation results demonstrate that the proposed method can obtain comparable results to other state-of-the-art approaches. The proposed method achieves the overall sensitivity of Sen = 99.64%, positive predictivity of PPR = 99.71%, and detection error rate of DER = 0.19.