A new filtering method for smoothing intracardiac records preserving the steepness of A, V, H waves

Oto Janousek1, Jakub Hejc2, David Pospisil3
1Department of Biomedical Engineering, Brno University of Technology, Brno, Czech Republic, 2International Clinical Research Centre (ICRC), St. Anne's University Hospital, Brno, Czech Republic & Department of Pediatrics, Children's Hospital, University Hospital Brno, Brno, Czech Republic, 3Department of Internal Medicine and Cardiology, University Hospital Brno and Faculty of Medicine of Masaryk University Brno, Czech Republic


Aim: Intracardiac electrocardiogram (IECG) based diagnosis requires that the IECG be legible, that is, not lost in noise and not morphologically disturbed. However, due to low voltage levels, IECG is often disturbed by electromagnetic noise during acquisition. Therefore, it is necessary to filter raw IECG records, which in turn undesirably changes the morphology of the IECG. The new filtering method introduced in this article can solve both at once.

Method: The method uses empirical mode decomposition. The principle of the method is the decomposition of IECG into a set of intrinsic mode functions, the subsequent identification of the intrinsic mode function containing the noise component and the use of its parameters for modeling the oscillating component, which is subtracted from the original IECG. Then the whole process is repeated several times, which removes each of the harmonic components of the noise. The process is terminated when all harmonic components – that have been identified by spectral analysis prior to the start of the process – have been removed.

Results and conclusion: The described method for smoothing IECG records preserves the steepness of rapidly changing sections of IECG, such as A, V, and H waves, and at the same time smooths the rest of the record. It has been proven on clinical adult IECG dataset, that the method removes noise even at high noise levels. Unlike other types of filtration – narrow band filtering introduces ripples into the signal, averaging techniques devalue the sharpness of A, V, H waves – proposed method does not change the morphology of the signal, thus enabling both visual diagnosis of cardiac disorders and automatic IECG classification that would not be feasible in the native record.