A flexible PCA-based ECG-reconstruction algorithm with confidence estimation for ECG during exercise

Holter monitoring of the electrical activity of the heart is the key for analysis of many intermittent cardiac irregularities. The measurements are however prone to movement artifacts or poor skin-electrode contact, often making some parts of the recorded signals useless since important information might be hidden by these artifacts. In this paper a special ECG sensor node with active electrodes is used to record the 12-channel ECG during movement. Several methods for detection of degraded signal quality are combined and adapted to identify leads that have been compromised on a beat-to-beat level. A PCA-based reconstruction algorithm is then used to reconstruct leads with insufficient signal quality using only undisturbed leads. Furthermore this approach enables an estimation of the quality of reconstruction even in absence of an undisturbed reference.