The unpredictable nature of epileptic seizures places people with epilepsy under permanent psychological stress, which contributes significantly to a restriction in their quality of life. The possibility of predicting the arrival of epileptic seizures would allow, in addition to taking a preventive treatment if the risk of seizure is close, to prevent traumas and accidents linked to possible falls during seizures, to authorize driving for certain people with epilepsy and to reduce the costs of medical care. To date and to our knowledge, no seizure detection device has been commercialized. There are commercialized devices based on biometric sensors other than EEG, but these are strictly dedicated to the detection of seizures and do not allow the anticipation of seizures. Regarding prediction, current research seems to have difficulties in developing convincing algorithms. The only system used successfully in real time would require a device implantable in the brain, but this would raise problems of acceptability.

In addition, 20% of people with drug-resistant epilepsy have psychogenic non-epileptic seizures (PNES). These are sometimes difficult to differentiate from epileptic seizures by people with epilepsy and their caregivers, and their management differs from that of epileptic seizures. The distinction between these 2 types of events should also be taken into account by these prediction/detection tools.

From the field of biomedical detection dogs, there is currently a converging body of evidence supporting that people with epilepsy emit specific odors associated with seizure events. Trained dogs have been shown to be able to discriminate body odors sampled during or just after an epileptic seizure from those sampled from the same subjects in various contexts outside of a seizure. It was also shown that a seizure can also be predicted by the volatile organic compounds (VOCs) released by the patient (human volatilome); the olfactory signature being already detectable up to 3h before a seizure. Another study used trained dogs to confirm that they are able to detect a seizure by smell and that this olfactory difference is already detectable before a seizure.

The human volatilome VOCs lead is particularly promising, notably for its non-invasiveness and for the pre-ictal precocity that prediction allows. But at the moment, the studies are too studies are too preliminary, with sample sizes too small to conclude on the inter-individual generalization of the odor, taking into account the type of seizure involved and the influence of other variables (e.g., gender, age, medications). Moreover, in order to develop a reliable and transportable electronic detection tool, the identification of the VOCs involved is necessary, since the choice of sensors (e.g., to constitute an electronic nose) depends on it. The objective of this study is to overcome these shortcomings, by aiming at the identification of the informative odor(s) associated with epileptic events during the pre-ictal, ictal and post-ictal periods, taking into account the type of seizures (focal seizures, secondary generalized focal seizures, primary generalized seizures - motor and non-motor) and the inter-individual differences.

Inclusion Criteria:

• Person at least 18 years of age
• With drug-resistant epilepsy according to the ILAE criteria
• With one of the following 3 types of seizures :

Focal seizures Focal seizures with secondary generalization Generalized seizures - motor or non-motor

• Requiring at least 48 hours of video-EEG
• Consent to participate in the study from the patient and his/her legal guardian, if applicable
• Affiliated or beneficiary of a social security plan

Exclusion Criteria:

• Person with epilepsy benefiting from a legal protection measure other than curatorship or guardianship
• Pregnant or breastfeeding woman

The epileptics will be recruited by the neurologists from among the epileptics treated at the Institut la Teppe and the Fondation A de Rothschild Hospital who meet the selection criteria.