EFFECT OF ECG REMOVAL ON LUMBAR EMG HALF-WIDTH

Introduction

It has previously been reported ^(1,2,3) that EMG signals from the lumbar spine are highly prone to contamination by ECG artefacts. It has also been reported that Independent Component Analysis is a suitable method for extracting this contamination (4).

Methods

EMG data was recorded from 192 subjects across two years (initial contact, 12 months and 24 months). The data were analysed and the spectral half-widths calculated.

The ICA method was then applied to the original raw data. As the power spectrum of ECG runs from 0-20Hz the resultant spectra were analysed to calculate which of them had the most signal energy below 20Hz. A high band pass filter was used to remove all signal data below 20Hz from this independent component.

This method was chosen as there was signal data present in the chosen spectrum above 20Hz which would be EMG data. Removing data only below 20Hz preserved this EMG data.

The components were then re-integrated and re-analysed to calculate the new half-widths. These new half-widths were compared with the originals to generate the results.

Results

Originally the half-width variable distinguished between back pain suffers and non-back pain suffers with a sensitivity of 13.2% and a specificity of 76.5%. However after the ICA algorithms were applied the sensitivity became 30.0% and the specificity 58.6%. This means that while this technique is effective at removing ECG contamination from lumbar EMGs specifically identified as contaminated, the blind application of this technique cannot be recommended.