Artefact Detection in Sleep Eeg by the Use of Kalman Filtering

Inverse filtering can be used to identify transient events. Often, artefacts in the EEG are such transient events. Sleep EEG data of 8 eight European sleep labs were scored in 1s-epochs for 9 types of artefacts. The area under the ROC curve (AUC) was 0.857 and 0.898 for muscle and movement artefacts, respectively. Kalman filtering can be used to estimate adaptive autoregressive (AAR) parameters and apply adaptive inverse filtering, simultaneously. The variance of the prediction error can be used as a indicator for muscle and movement artefacts.