Implementation of different techniques for removing artifacts in diaphragmatic SEMG, and evaluation of these in automatic and online applications

Abstract

Respiratory muscles superficial electromyography (SEMG) is an important source of information in the monitoring of ventilated patients. One of the main problems in the acquisition of SEMG signals is the different sources of interference. The most common artifacts are the baseline wander (BW) normally generated by motion, and power line interference (PLI). In this paper, different methods were selected and evaluated for the removal of these artifacts in a simulated SEMG signal of the right diaphragm muscle. The best performance technique for the removal of each artifact was determined using frequency analysis and estimation of criteria such as the signal to noise ratio, relative error, cross-correlation, and coherence of the power spectrum density. The computational cost of each of the techniques was estimated to also assess how appropriate it is to implement in online applications and limited hardware. The study demonstrates that the spectral interpolation technique has a good performance in removing PLI from the SEMG signal but has a high computational cost, unlike the adaptive LMS filter. On the other hand, the SSA-based technique proved to be the best performing for BW removal and its computational cost is adequate in a more limited hardware system.