Multimodal information fusion for robust heart beat detection

QRS detection based on ECG signal is the most straightforward method for heart beat detection. However, existing QRS detection methods do not work well when ECG signal is contaminated or missing. Other physiological signals also contain information about cardiac activity and ECG. Their information can be explored for robust heart beat detection. As part of the PhysioNet/Computing in Cardiology Challenge 2014, this study proposed a multimodal information fusion framework for robust heart beat detection. The framework consisted of three steps: 1) QRS detection. 2) Remove spurious QRS detection using pulsatile signal if it is available. 3) Refine the remaining beat detection and interpolate missed beats. Results show that the algorithm can sufficiently reduce spurious QRS detection and accurately fill in missed beats.