EEG source imaging concordance with intracranial EEG and epileptologist review in focal epilepsy

Abstract

EEG source imaging is becoming widely used for the evaluation of medically refractory focal epilepsy. The validity of EEG source imaging has been established in several studies comparing source imaging to the surgical resection cavity and subsequent seizure freedom. We present a cohort of 87 patients and compare EEG source imaging of both ictal and interictal scalp EEG to the seizure onset zone on intracranial EEG. Concordance of EEG source imaging with intracranial EEG was determined on a sublobar level and was quantified by measuring the distance between the source imaging result and the centroid of the active seizure onset zone electrodes. The EEG source imaging results of a subgroup of 26 patients with high density 76-channel EEG were compared with the localization of three experienced epileptologists. Of 87 patients, 95% had at least one analysis concordant with intracranial EEG and 74% had complete concordance. There was a higher rate of complete concordance in temporal lobe epilepsy compared to extratemporal (89.3 and 62.8%, respectively, P = 0.015). Of the total 282 analyses performed on this cohort, higher concordance was also seen in temporal discharges (95%) compared to extratemporal (77%) (P = 0.0012), but no difference was seen comparing high-density EEG with standard (32-channel) EEG. Subgroup analysis of ictal waveforms showed greater concordance for ictal spiking, compared with rhythmic activity, paroxysmal fast activity, or obscured onset. Median distances from the dipole and maximum distributed source to a centroid of seizure onset zone electrodes were 30.0 and 32.5 mm, respectively, and the median distances from dipole and maximum distributed source to nearest seizure onset zone electrode were 22.8 and 21.7, respectively. There were significantly shorter distances in ictal spiking. There were shorter distances in patients with Engel Class 1 outcome from surgical resection compared to patients with worse outcomes. For the subgroup of 26 high-density EEG patients, EEG source localization had a significantly higher concordance (92% versus 65%), sensitivity (57% versus 35%) and positive predictive value (60% versus 36%) compared with epileptologist localization. Our study demonstrates good concordance between ictal and interictal source imaging and intracranial EEG. Temporal lobe discharges have higher concordance rates than extratemporal discharges. Importantly, this study shows that source imaging has greater agreement with intracranial EEG than visual review alone, supporting its role in surgical planning.

Keywords: EEG source imaging; epilepsy surgery; intracranial EEG; source localization.

Graphical Abstract

Figure 1

ESI results of a non-lesional patient with standard density EEG. ESI dipole (green), intracranial electrodes (black, blue for seizure onset zone) and centroid of seizure onset zone (magenta cross). Sublobar localization of intracranial EEG was left insula. Resection was not performed due to eloquent cortex, but patient has been seizure-free for 2 years with chronic subthreshold stimulation in left insula. (A) initial 2 s of ictal onset. sLORETA sublobar localization was left insula, left mesial frontal. Dipole-centroid distance (using fixed MUSIC) was 18.1 mm and to nearest SOZ electrode was 10.9. MSL-centroid distance was 19.5 mm and to nearest SOZ electrode was 12.2 mm. SNR 1.38. (B) Average of 14 interictal broad sharp waves, non-propagating. sLORETA sublobar localization was left insula and left temporopolar. Dipole centroid distance was 15.3 mm and to nearest SOZ electrode was 15.2 mm. MSL-centroid distance was 28.1 mm and to nearest SOZ electrode was 21.4 mm. SNR 1.90. (C) Average of 32 sharp waves, propagating. sLORETA sublobar localization was left mesial temporal, anterior, temporal and insula. Dipole-centroid distance was 23.5 mm and to nearest SOZ electrode was 16.6 mm. MSL-centroid distance was 23.8 mm and to nearest SOZ electrode was 19.5 mm. SNR1.8.

Figure 2

ESI results of ictal and interictal standard density EEG in a non-lesional patient with subdural grids and strips. ESI dipole (green), intracranial electrodes (black, blue for seizure onset zone) and centroid of seizure onset zone (magenta cross). Sublobar localization of icEEG was right mesial temporal, right temporopolar. The patient was Engel class 1A at 1-year follow-up after anterior temporal lobectomy. (A) Average of 14 interictal sharps, propagating. sLORETA sublobar localization was right mesial temporal, right temporopolar. Dipole-centroid distance was 8.3 mm and to nearest SOZ electrode was 6.7 mm. MSL-centroid distance was 12.4 mm and to nearest SOZ electrode was 6.5 mm. SNR 4.5. (B) Initial 2 s of ictal onset. sLORETA sublobar localization was R mesial temporal, insula, inferior frontal. Dipole-centroid distance (using fixed MUSIC dipole) was 27.7 mm and to nearest SOZ electrode was 16.8. MSL-centroid distance was 29.4 mm and to nearest SOZ electrode was 19.7 mm. SNR 2.9.

Figure 3

ESI results of high-density EEG in a non-lesional patient with sEEG. ESI dipole (green), intracranial electrodes (black, blue for seizure onset zone) and centroid of seizure onset zone (magenta cross). Sublobar localization of icEEG was right mesial temporal, right temporopolar. Sublobar localization of epileptologist was right mesial frontal and superior frontal (blue shading). The patient did not undergo resection due to overlap with eloquent cortex. sLORETA localization was right mesial frontal. Dipole-centroid distance (using fixed MUSIC) was 15.6 mm and to nearest SOZ electrode was 13.0 mm. MSL-centroid distance was 17.2 mm and to nearest SOZ electrode was 15.2 mm. SNR 1.4.

Figure 4

ESI results of ictal and interictal high-density EEG in a non-lesional patient with sEEG. ESI dipole (green), intracranial electrodes (black, blue for seizure onset zone) and centroid of seizure onset zone (magenta cross). Sublobar localization of icEEG was independent right and left mesial temporal SOZ (multifocal). The patient enrolled in an investigational brain stimulation trial. (A) Average of 14 propagating sharp waves. sLORETA localization was left mesial temporal, temporal polar. SNR 4.3. (B) Initial 8 s of seizure onset. sLORETA sublobar localization was left mesial temporal, temporopolar. SNR 1.82. (C) Initial 5 s of ictal onset. sLORETA localization was right mesial temporal. SNR 2.4.

Figure 5

ESI results of ictal and interictal standard EEG in a patient with prior right frontal resection and subsequent sEEG. ESI dipole (green), intracranial electrodes (black, blue for seizure onset zone) and centroid of seizure onset zone (magenta cross). Sublobar localization of icEEG was R frontal superior, frontal mesial. Patient underwent additional resection and was Engel class 3A at 1-year follow-up. (A) Average 10 interictal sharp waves, non-propagating. sLORETA localization was right superior frontal and mesial frontal. Dipole-centroid distance was 18.9 mm and to nearest SOZ electrode was 6.3 mm. MSL-centroid distance was 16.9 mm and to nearest SOZ electrode was 5.2 mm. SNR 2.2. (B) Initial 5 s of seizure onset. sLORETA localization was right mesial frontal and inferior frontal. Dipole-centroid distance was 29.1 mm and to nearest SOZ electrode was 13.0 mm. MSL-centroid distance was 16.1 mm and to nearest SOZ electrode was 8.2 mm. SNR 3.4.