Ultra long-term subcutaneous EEG monitoring of brain functioning and disease

Both seizures and the fear of having another seizure can have far-reaching consequences for patients with epilepsy. In many cases, anti-epileptic drugs can provide a better seizure control and thereby a better control of one’s everyday life. Epileptologists must balance the pros and cons of each treatment option when choosing the best care for their patients. Poor sleep can lead to a higher seizure propensity, but seizures and anti-epileptic drugs can also negatively affect sleep quality. Seizures, drugs and sleep all impact the patient’s quality of life. There is therefore a growing consensus that possible sleep disturbances should also be taken into account when comparing treatment schemes.The interaction between epilepsy and sleep may vary depending on the type of epilepsy and other factors specific to a given patient. The patient specific interaction could be illuminated by observing how seizure occurrence and sleep quality evolve over extended periods of time. Monitoring sleep and seizures by technological means could reveal patterns unknown to the patient, as subjective accounts of these are known to be inaccurate. Objective seizure and sleep measures could therefore provide crucial information to the treating epileptologist. Continued monitoring could also reveal changes in seizure patterns in response to altered medication. This thesis presents a pilot study testing a wearable, subcutaneous EEG device on pa-tients with epilepsy. Nine patients wore the device continuously for three months. Fourof the nine patients furthermore wore the device while submitted to the epilepsy monitoring unit, providing concurrent scalp and subcutaneous EEG measurements. Basedon the concurrent recordings, we show that epileptiform and sleep related events appear similar in the subcutaneous recordings compared to scalp EEG recorded from similar electrode positions. Furthermore, by training patient-specific sleep staging algorithms, we obtain classification performances on par with human experts. Based on work performed on a publicly available EEG database, we show that there is a potential for saving time and resources by training personalized sleep models in a semi-supervised fashion.Based on data collected while the patients wore the device in their everyday lives for three months, we show that the device is safe to use and is well-tolerated by the patients. A comparison of seizure diaries and seizures visually detected in the subcutaneous EEG revealed that the self-reported seizure pattern often differed from the electrographical seizure pattern. We cannot know whether the discrepancy stems from undetected seizures in the EEG, or whether the patients suffers from seizure unawareness. As in-formation on seizure occurrence is important for the treating epileptologist, we argue that the additional information provided by at-home-monitoring could help make better-informed treatment decisions.A small investigation of how sleep length affects the probability of having a seizure the next day, revealed that the collected dataset was not optimal to statistically investigate this complex question. A qualitative analysis revealed that severe sleep deprivation did influence seizure propensity, but that there were other, unknown factors influencing the seizure patterns as well. Without more information about these other factors (that may or may not interact with sleep), it is difficult to tease out the predictive power of sleep length alone. Another small study investigated whether it was feasible to monitor citizens of a Danish ”tech town” using a wrist-worn activity tracker that measures the amount of physical activity and sleep for four weeks. The study was conducted as preparation for a larger upcoming study that aims to investigate how using an activity tracker affects aerobic capacity and how removing electronic media devices from the bedroom affects sleep. The results showed that compliance of wearing the activity tracker was high, and that sleep estimates were obtained for most nights. The participants using the activity tracker for fitness monitoring increased their aerobic capacity and reduced their systolic bloodpressure. The experience from this small feasibility study suggested a few adjustments that will be incorporated in the future, large-scale study. Although this thesis shows that it is feasible to monitor patients for long periods ”in the wild”, future research is needed to illuminate the clinical utility of this new treatment aid.