The distance between the estimated global max dipoles and the other three local maxi-mums is, for MLE, 118.3 mm to the max dipole in BA 4, 10 mm to the max dipole in BA 7, and 73. mm to the max dipole in BA 30. For MNE the distance between the local maximums and the estimated global max is 118.3 mm to the one in BA 4, 10 mm to the one in BA 7, and 73.7 to the one in BA 30. And for wMNE the distances are 5 mm to the one in BA 4, 115.9 mm to the one in BA 7, and 71.8 mm to the one in BA 30. The second greatest con-tributor to the EEG was the dipole located in BA 7, which was the one that MLE and MNE got closer to. The wMNE got closer to the local max dipole located in BA 4.
SN R = −4, the highest level of noise. It is expected that with the decreasing of noise in the signal, better estimations could be achieved.
Figure 3.14: Case five: EEG-SL with SN R = −4. The left column shows the dipole estima-tion done with MLE, the middle column shows the estimaestima-tion done with MNE and the right column shows the estimation done with wMNE.
Figure 3.15: Case five: EEG resulting from the SL with SN R = −4. The left column shows the EEG produced by the estimation done with MLE, the middle column shows the EEG produced with the estimation done with MNE and the right column shows the EEG corresponding to the SL done with wMNE.
Figure 3.16: Case five: EEG-SL with a zero SNR. The left corresponds to the SL done with MLE, the middle column shows the SL done with MNE and the right column shows the SL done with wMNE.
The results of SL with SN R = −4 can be seen in fig. 3.13 with the respective EEG in fig. 3.15. The estimation done with MLE is in the left, with MNE in the middle and with wMNE in the right. As expected, from the brain activity it’s quite difficult to discern a main
Figure 3.17: Case five: EEG resulting of the SL with a zero SNR. The left corresponds to MLE, the middle to MNE and the right column for wMNE.
Figure 3.18: Case five: EEG-SL with SN R = 4. The left column shows the dipole estimation in each case with MLE, the middle column shows the estimation with MNE and the right column shows the estimation with wMNE.
Figure 3.19: Case five: EEG resulting from the SL with SN R = 4. The left column shows the EEG in each case produced by the estimation of MLE, the middle column shows the EEG for MNE and the right column shows the EEG for wMNE.
source. By observing the EEGs it is notable that the three results are very similar as the original EEG.
The results of the SL for the values of SN R = −3, SN R = −2 and SN R = −1 are displayed in fig. A.12 and the corresponding EEG is found in fig. A.13, in the Appendix.
The results for when SN R = 0 are shown in fig. 3.16 and fig. 3.17. It is still difficult to see with a simple visual inspection if there is a main source. Furthermore, it can be appreciated in the colorbars that the activity is diminishing with respect with the lower SN R. This is also observed in the EEG images.
The reader can observe the results for the scenarios with SN R = 1, SN R = 2 and
Figure 3.20: Case five metrics: comparison between methods for case five.
SN R = 3 in fig. A.14 and their respective EEG in fig. A.15, in the Appendix.
For the last scenario with SN R = 4 the resulting SL is shown in fig. 3.18 and the corresponding EEG is shown in fig. 3.19. In this last scenario it still difficult to distinguish the main source regardless of the SLM used. By observing the colorbar we can clearly see how the activity has been dimmed. By examining the EEGs the same EEG as in the previous cases without noise can finally be observed.
The quantitative metrics plotted in fig. 3.20 give a more precise performance of each SLMs. The top left plot shows the evolution of the distance metric throughout the case. In this plot it is observed that regardless of the SNR every SLM obtained the same distance from the original source in their estimations. The top right plot shows the goodness of fit. This plot demonstrates how the estimated EEGs are almost equal to the original EEG regardless of the SNR. The middle plots show the metrics for the maximum dipole estimated compared to the original maximum dipole. These plots show the errors reaching towards zero as the SNR grows. They also show almost no difference between SLMs. The bottom two plots
show the metrics regarding every dipole in the brain. As in the previous two plots they show the errors dropping to zero as the SNR grows larger. These metrics corroborate the initial assumption, the less noise is present in the signal, the better the estimation obtained. However, it is important to note how good has been the goodness it even with a lot of noise present in the signal.