• stereo-electroencephalography
• seizure pattern
• focal epilepsy
• epilepsy surgery

Background. Despite a great number of antiepileptic drugs available, one-third of epilepsies remains medically intractable. A safe and effective treatment option for focal pharmacoresistant epilepsy is the surgical resection of the cortical area responsible for seizures generation and their primary organization (epileptogenic zone, EZ). The identification of the EZ is fundamental for a successful surgery and is based on the anatomo-electro-clinical correlation established through seizures observation and electroencephalographic (EEG) recordings. Imaging techniques also play a role in this process, and invasive studies may be required for the most complex cases. Stereo-electroencephalography (SEEG) allows prolonged (1-3 weeks) recordings with intracerebral electrodes, aimed at localizing the EZ during the recording of spontaneous seizures. After the SEEG, a resection is usually proposed, but in some cases a valid alternative is radiofrequency thermocoagulation (THC) of the EZ through the same recording electrodes. The outcome of epilepsy surgery depends on the correct EZ identification and is evaluated based on seizure recurrence. The visual analysis of SEEG signal of a single patient by expert epileptologists requires several days. To assist and expedite this process, in the last decade computerized methods for SEEG analysis were developed.

Material and methods. In this study we retrospectively analyze the SEEG seizures recordings from 176 patients with a computer-assisted method developed by Gnatkovsky et al. with the Elpho-EEG™ software. The spectral analysis of intracranial recordings allows the identification of three different focal seizure patterns and the localization of the EZ with specific frequency peaks at seizure initiation (fast activity and slow polarizing shift).

Results. We further characterize three specific pattern subtypes, defined as P-, L-, and P+L-type, based on seizure discharge duration and involvement of the mesial temporal regions. P-type seizures present with an abrupt and large amplitude slow polarizing shift (SPS), with sharp on and off transients and a superimposed low-voltage fast activity (LVFA) at 100-200 Hz, that linearly decreases in frequency (down chirp) in 5-10 s. The down chirp was described by other authors as a fingerprint spectrographic signature of seizure onset in SEEG. Mean seizure duration is 26 ± 20 s. L-type seizures start with LVFA at 116 ± 21 Hz) superimposed on a slowly developing SPS; they evolve in tonic spiking discharge and terminate with a bursting phase. The duration is longer than P-type seizures, with a mean of 112 ± 47 s. L-type pattern was found more often in the temporal lobe, associated with different pathologies. P+L-type pattern presents with a short (mean: 7.8 s) P phase followed by an L phase, and has a total duration of 81 ± 46 s. It often secondarily involves mesial temporal regions. The three patterns have similar seizure outcome after surgery or THC. They all can reflect different underlining neuropathology, but a significative association was found between L-type seizure pattern and hippocampal sclerosis.

Conclusions. The method we applied show high operator-independent reproducibility and allows quick identification of probable seizure-onset areas with no information on clinical features, imaging findings, or electrodes’ position. We utilized the computer-assisted method and the traditional visual assessment to validate the comparison between the two methods and to gather information for an easier and more objective EZ detection.