Comparison of spatial filter selectivity in surface myoelectric signal detection: Influence of the volume conductor model

Spatial filters are used for increasing selectivity in surface EMG signal detection. The study investigated the importance of the description of the volume conductor to the inference of conclusions on comparing filter selectivity, from simulation analyses. A cylindrical multi-layer description of the volume conductor was used for the simulation analysis. Different anatomies were analysed with this model, and results on filter selectivity were compared. The longitudinal single (LSD), double (LDD) and normal double differential (Laplacian, NDD) filters were investigated. Largely different conclusions could be drawn when comparing filter selectivity resulting from simulations with different volume conductor models. A filter that performed best with a particular anatomy could be the poorest with another anatomy. With a bone-muscle model and superficial fibres, the ratio between peak-to-peak values of the propagating and non-propagating signal components was approximately 220% for LDD and LSD and lower than for NDD (approximately 290%). With a bone-muscle-fat-skin model, LSD performed significantly worse (150%) than both LDD and NDD, which showed similar performances (approximately 300%). Similarly, if the lateral distance of the recording was increased by 10°, the signal amplitude was reduced to 2% with LSD and LDD and to 4% with NDD. With another anatomy, LSD and LDD reduced signal amplitude to 20–25%, and NDD reduced it to 4%. Similar considerations could be drawn for other selectivity indexes. Thus, modelling should be used carefully to infer conclusions on spatial selectivity and to indicate particular choices of spatial filters.