Abstracts

Rationale: 30-40% of patients with epilepsy are refractory to medication, and further surgical treatment requires identification of the seizure onset zone (SOZ). Imaging in the postictal period is underexplored. Prior work (Farrell JS, et al. eLife 2016;5:e19352) showed that stimulated seizures in rats result in severe hypoxia in the SOZ for approximately 60 minutes. This suggests that postictal imaging of hypoperfusion may localize a SOZ. Arterial spin labeling (ASL) is an MRI perfusion technique that quantifies cerebral blood flow (CBF). Postictal subtraction ASL (interictal baseline CBF minus postictal CBF) is a novel application. A prior study (Gaxiola-Valdez I, et. Brain 2017;140:2895-2911) reported hypoperfusion in 71% of cases, and in 80%, the area of maximal hypoperfusion corresponded to the SOZ. Our methodology improves on previous work.


Methods: Subjects with refractory focal epilepsy were recruited for ASL at two time points during their Epilepsy Monitoring Unit (EMU) admission at the University of Kentucky: interictal baseline ASL at admission and postictal ASL 20-60 minutes after seizure offset. Subjects were scanned on a 3T MRI scanner using 3D pseudo-continuous ASL sequence with multi-TI technique that uses multiple time delays to accurately calculate arterial transit time. To reduce noise, we used a smoothening strategy that replaces a voxel’s values by the average of its neighboring voxels, and subtraction ASL was thresholded to 1% maximum hypoperfusion. The ground truth of SOZ localization was determined at two levels of confidence: level 1, expert consensus determined by presurgical evaluation, and level 2, postoperative stereo-EEG or responsive neurostimulation (RNS) results. Concordance of ASL hypoperfusion was determined at these two levels of confidence, and postictal ASL was compared to ictal SPECT as both modalities are time-sensitive dynamic scans around the seizure.


Results: We recruited 11 subjects. Mean seizure-to-scan delay was 37 minutes. (In Gaxiola-Valdez’s study, half were >60 min.) Applying confidence level 1 at lobar resolution, ASL hypoperfusion was concordant in 10/11 (91%) subjects. Applying confidence level 2 at gyral resolution, ASL hypoperfusion was concordant in 2/3 (67%) (Figure 1 & 2). Ictal SPECT was obtained in 6/10 (75%) eligible seizures, while postictal ASL was done in 9/10 (90%). Ictal SPECT was concordant to the correct lobe in 3/10 (30%) seizures, while ASL succeeded in 7/10 (70%). However, postictal ASL is burdened by false positives, each case having an average of 8 hypoperfusion areas. One subject had hyperperfusion in the SOZ, which shows that ongoing seizures can mask hypoperfusion. Considering either hypoperfusion or hyperperfusion, ASL is concordant in 100% of subjects at lobar resolution, and 100% at gyral resolution.


Conclusions: Postictal subtraction ASL hypoperfusion is concordant to the correct lobe in 91% subjects and to the correct gyrus in 67%. Preliminary results are promising, but a methodology is necessary to reduce false positives & false negatives.


Funding: Supported by the University of Kentucky College of Medicine through Alliance Initiative & NEUSTAR awards.