An Algorithm for Removing Artifacts in Polysomnography Signals

Polysomnography (PSG) is considered the gold standard for sleep disorders diagnosis. However, its signals are difficult to read in the presence of artifacts. This study presents a biosignal processing method that can help to remove artifacts from PSG signals. The method consists of two main parts. Firstly, sleep diary and feature extraction using Fourier transform and Wavelet transforms were used to detect different artifacts in PSG signals including physiologic artifacts such as cardiac artifact, muscle artifact, movement artifact, ocular artifact...and non-physiologic artifacts. Secondly, ICA (Independent Component Analysis) and Wavelet transform were used to remove the detected artifacts and reconstruct the signal. The results indicate that our method could adequately remove ECG and EOG artifacts in 70%–80% of the cases and wandering baseline and 50 Hz artifacts in 100% of the cases. We hope this algorithm will become a useful tool in removing artifacts in PSG signals, thus helping to make PSG signal reading more straightforward.