Sentence examples for maximum localisation error from inspiring English sources
GN shows markedly inferior localisation performance to the other two algorithms at all noise levels and for both sources, with a maximum localisation error of 5.3mm.
NNLS shows similar performance for source A, but consistently inferior performance for source B, with a maximum localisation error of 1.7mm as compared to 1.1 mm for CSCG.
Examining reconstruction performance quantitatively, Table 1 shows that CSCG displays superior localisation performance to both GN and NNLS, with localisation better or equal to the other algorithms at all noise levels and for both sources, with a maximum localisation error of 1.7mm.
In cases where one CPD-based dipole is slightly better localised than the BTD-based one, the localisation error of the second dipole is much worse.
To reduce the impact of localisation error on these results a temporal Gaussian smoothing filter of FWHM 4 frames (200 ms) was applied to the position traces before the colocalisation analyses.
Subsequently, one dipole was fitted on the extracted ictal source signal to compute the localisation error.
The goodness of the decomposition was evaluated in terms of the dipole localisation error.
However, results obtained from the second CPD component are omitted in this figure, as the corresponding localisation error exceeded 10 cm.
(a) Localisation error of the dipole fitted on the reconstructed EEG based on a single CPD component and the two dipoles fitted on the reconstructed EEG based on BTD for various SNR values.
Regarding the retrieval of the spatial maps, all three BSS approaches performed equally well, reaching an RMSE in spatial distribution below 0.6 with the simulated ictal source, which corresponds to a dipole localisation error of less than 5 mm.
Alternatively, one can take the sum of these two reconstructed EEGs and two dipoles can be fitted on the resulting signal.The localisation error of the extracted sources with respect to the corresponding simulated source is shown on Figure 12.
