Removing noise from electroencephalogram signals for BIS based depth of anaesthesia monitors

Removing noise from electroencephalogram signals for
BIS based depth of anaesthesia monitors

The assessment of patient has changed from the physical assessment to digital assessment. One significant example is the assessment of the depth of anaesthesia (DoA). It has changed from physical to digital assessment using DoA monitor. DoA monitor uses the electroencephalogram (EEG) signal as its input. The processes include the digitising, filtering and signal analysing. This study focuses on filtering process to reduce noise in the EEG signal.

Noises in EEG signals could affect the accuracy of DoA monitor. The noises in EEG signal are from the muscle, eye movement and blinking, power line, and interference
with other device. Those noises are overlapped each other. Hence, monitoring of DoA without removing the noise may result in an incorrect assessment. A simple filtering process such as band pass filter is not able to remove all noise from EEG signals.

There are three methods which are introduced to remove noise from EEG signals. The first technique is adaptive least mean square technique, which is able to find the
best output of the signal through the iteration. In this method ANOVA technique is employed to define the best coefficient of the signal in the adaptation. This technique
is chosen to find the significant output from the iteration. The result shows that the adaptive least mean square with the ANOVA is able to remove the noise from EEG
signal effectively.

second method is Wavelet transform. In this technique, EEG signal is decomposed into five levels using the Stationary Wavelet Transform (SWT). The first step of this filtering is to eliminate high frequency noise in the EEG signal. The
next step is to remove the low frequency noise using the soft threshold method. The final step is to reconstruct the signal. The result from this method shows that there is
a significant improvement of the signal quality after the filtering.

The third method is a combination of adaptive LMS and wavelet transforms method. The result from this study shows that the wavelet transform adaptive filter is able to
remove both the low frequency noise and high frequency noise in the EEG signal. Compare to the previous two other methods, the combined method is also more robust. Filtering the noise in EEG signal with wavelet transform adaptive filter technique could minimise false prediction of DoA.