The use of different measures of signal shape for automatic identification of artefacts in impedance cardiography

The aim of this work was to find an easy and efficient method for automatic detection of artefacts in ICG signals. Form factors characterising the shape of the signal were selected because of the simplicity of implementation and the low computational cost. Different form factors were used to compare a single ICG heart cycle with a pattern and classify it as valid or artificial. The pattern was made by averaging 50 consecutive evolutions synchronized by the Q wave in an accompanying ECG signal. If the absolute difference between values of the same form factor calculated for the single evolution and for the pattern is lower than the cutoff point, it is recognized as valid; otherwise, it is marked as artefact. The main objective of the study was to choose the most effective of the commonly-used form factors. Effectiveness was determined by using the area under the curve in receiver operating characteristic analysis. The necessary data were obtained by analysing the absolute difference between the values of the same form factor calculated for a single, manually classified evolution and for the pattern. We analysed cycles produced by 5 minutes of impedance cardiography observation in each of 20 subjects. The best efficiency was identified for the normalized standard deviation of the cycle from the pattern. In this case, area under the curve (AUC) was 0.86 and the cost-effective cut-off point was 56.76.