Abstracts

Rationale: The quality of life (QOL) of women with epilepsy may be improved by managing antiseizure medication dose based on forecasted seizure risk thus reducing the effects of seizures and medication on menstrual cycle, pregnancy, menopause, bone health, and reproductive health. Analysis of physiological data recorded with non-invasive wearable sensors and implementation of machine learning methods allows a better understanding of the impact of physiological hormonal rhythms on seizure forecasting in women of childbearing age. The accuracy of seizure forecasting algorithms could be significantly improved if the seizure risk associated with hormonal changes be considered in the algorithm design.

Methods: Previous reports confirm that the temporal distribution of seizures is non-random, and the risk of seizure occurrence can be forecasted. Hormonal changes influence brain excitability and can modulate seizure risk. Luckily the hormonal changes in women are associated with measurable changes in bio-signals such as temperature and heart rate variability. Therefore, correlating hormonal changes and seizure pattern is achievable by using bio-signals measured from non-invasive devices while tracking the menstrual cycle. Long-term physiological ambulatory data acquired from patients with epilepsy enrolled in the “My Seizure Gauge” project with simultaneous recordings via an implanted EEG device (Neuropace RNS) was used in this study. Seizures were annotated and confirmed by a board-certified epileptologist. Currently, we are collecting data from 15 healthy subjects to investigate physiological changes during menstrual cycles and we have long-term ambulatory data from 4 female subjects with epilepsy.

Results: Heart rate and temperature are affected by the phases of the menstrual cycle. In this study we confirmed that skin temperature and HR, measured during sleep using a wrist-worn device, show a biphasic pattern across the menstrual cycle, with increased HR and body temperature in the luteal phase relative to menstruation and ovulation. Correspondingly, HRV tended to be lower in the luteal phase. Long-term physiological ambulatory data collected from four epileptic female subjects shows the same pattern for two of them. The correlation between seizures and menstrual cycles is being investigated.

Conclusions: Designing a reliable seizure forecasting algorithm from data collected with noninvasive devices may allow patients to use lower baseline doses of medications, with escalated doses given during times of high seizure risk. Investigating the gender-specific parameters affecting seizure risk and implementing them in managing epilepsy in women results in significantly improved health-related quality of life in this patient population.

Funding: NSF-CBET # 2138378