Elise 

Epilepsy is one of the most prevalent neurological disorders, affecting around 52 million people who suffer from spontaneous and recurrent seizures due to abnormal electrical activity in their brains. In cases of drug-resistant epilepsy, the application of neurosurgical procedures represents a promising solution for seizure control. However, it requires precise localization of brain regions responsible for seizure origination, which is resource and time-consuming. Recordings of brain activity using stereo electroencephalograms (sEEG) are the primary tools used for seizure onset localization, which involves registering intracranial electrical activity with the help of electrodes surgically implanted into patients’ brains. Identifying seizure onset with sEEG recordings plays a critical role in surgery because it determines the initiation of pathological brain activity. Currently, this task is being performed by board-certified epileptologists, and it is extremely complicated and laborious. Despite numerous breakthroughs in AI-based seizure detection, little is known about its clinical relevance compared to the performance of medical experts. The purpose of this novel project is to train an unsupervised AI model to identify seizure onset on an sEEG recording with equal accuracy to board-certified epileptologists. To assess this, sEEG recordings of various seizures will be presented to board-certified epileptologists to identify the seizure onset. The AI will then be presented with the same recordings, and the data from both groups will be compared. Conclusions from this study may show that AI can be used to assist or partially automate the process of seizure onset recognition, increasing efficiency and consistency in surgery planning for patients with epilepsy.