It started with a question

“Which aspect of epilepsy impacts your life the most?”

The Epilepsy Foundation of America asked their community this question in a 2016 survey about their experience living with epilepsy. The most common answer was the unpredictability of seizures.

As a result of that survey, the Foundation put forward a fund of US$3m to create a solution to the unpredictability of seizures. The grant was awarded to Seer, Mayo Clinic, and King’s College London, to collaborate and develop a way to predict the likelihood of a person having a seizure. Our collective research efforts have led to the development of the world’s first non-invasive seizure cycle tracking technology — only available to research participants at this point in time.

In 2020, Seer was awarded an additional AU$1m by the Australian Government to support further development of the promising technology.

A different way to live with seizures

A seizure cycle tracker provides a new option for people to manage their epilepsy, particularly for people whose seizures are unresponsive to anti-seizure medications.

Approximately 30% of people with epilepsy are unable to control their seizures through the use of medication. The uncertainty of when a seizure will occur can put many people living with epilepsy in danger. Even small seizures can result in serious accidents if they happen whilst doing everyday activities like driving or cooking. Unfortunately, this means some people with epilepsy have to give up their driver’s licence, modify or avoid hobbies, and experience anxiety in their day-to-day lives.

Similar to how a weather forecast is helpful for making plans, a seizure cycle tracker can help people with epilepsy plan their activities. The ability to know when there is greater likelihood of a seizure occurring can help people prepare with confidence.

Turning decades of research into a solution

Seizure cycles are common for all people with any type of epilepsy. At Seer, our goal is to turn our knowledge of seizure cycles into a reliable method for forecasting seizure likelihood.

Since the earliest documented cases of epilepsy, people have reported cyclic rhythms to when and how their seizures occur. In ancient times, it was thought that the phases of the moon influenced seizure risk. Today, many people continue to report daily, weekly, or monthly events — which through mathematical analysis, reveal a pattern in their seizures.

Seer’s earlier research confirmed that seizures occur in cycles and that these cycles are common for all people with all types of epilepsy.

The cause of cycles remains a mystery and there are many possibilities, including internal and metabolic factors, changes in behaviour (such as sleep quality, stress levels, or exercise), and even environmental causes like seasonal variations. It’s likely that there are different causes for different people’s cycles.

A seizure cycle can exist without a person noticing. A major reason is that seizures can be infrequent. Another reason is that many people have multiple, overlapping cycles. A combination of cycles can appear to be irregular, but through specialised analytic techniques we can determine the cyclic rhythm and frequency of their seizures. 

Proportion of people with epilepsy who have seizure cycles in a study of 1,000 app users. Karoly et al (2018)

Measuring seizure likelihood

Using research participants events logged in the Seer app, we can determine seizure cycles and forecast seizure likelihood.

In our research, we measured the cycles of seizure risk and determined how reliable the cycles are to forecast someone’s chance of having a seizure. We used the seizure times recorded in the Seer app to detect two possible cycles: a ‘fast cycle’ and a ‘slow cycle’.

The fast cycle represents a short cycle of 1 or 2 days, likely driven by circadian rhythms of the body (e.g. sleep). The slow cycle could be one week or longer. Combining these two cycles develops a ‘score’ of seizure likelihood — low, medium, or high — that represents the risk of having a seizure in the next hour.

From these measurements, we can develop an individualised seizure cycle based on each app user’s logged events.

Making forecasts more powerful with wearable and implanted devices

Our ongoing research has established that pairing Seer app data with implanted or wearable devices, such as Epiminder and Fitbit respectively, can play an important role in forecasting seizure likelihood.

We’ve been able to improve the accuracy of seizure cycle tracking by partnering with Fitbit to gather other signals, particularly heart rate, which are very strong indicators of the cycles that take place inside the body.

The most accurate way to track someone’s seizure cycle is to continuously measure their brain activity, since it is ultimately the brain that causes seizures to occur. In fact, we know that long-term recordings of brain waves show very clear, multi-day cycles that are strongly predictive of seizure risk. Previously, these cycles have only been found using implanted brain recording devices.

However, we theorised that data from wearable devices, such as smartwatches, can be used to aid forecast models. In out study, we theorised that other organs, such as the heart, might also show long-term cycles related to seizure risk.

Our research participants wore a Fitbit smartwatch to record data on heart rate, stress, sleep and physical activity, and used the Seer app to record their event times. The results of the study found that wearable devices can provide powerful insights to improve seizure cycle tracking.

Personalised data is the key

Since seizure cycles are unique to a person, a personalised approach is critical to deliver the insights that people need. This process involves recording multiple types of data over long periods of time.

In order for the seizure cycle tracking feature to be as accurate as possible, multiple events need to be recorded. Typically we aim to capture at least 4 cycles (i.e. 4 weeks to detect a weekly cycle, 4 months to detect a monthly cycle) and approximately 10 events to initiate a forecast. In general, forecasts become more accurate as more events are recorded. However, our research team is actively working to reduce the time it takes to produce reliable forecasts of likelihood.

“Likelihood” reflects the underlying biological rhythms that impact seizure occurrence. Other factors within an individual’s control such as taking medications on time, stress, and sleep can also affect the likelihood of seizures.

Determining a seizure cycle can be as straightforward as looking at a list of past events and forming a forecast. However, a tailored seizure cycle tracker can only be delivered by looking at an individual’s seizure cycle(s) and taking additional factors into consideration.

Future considerations

Seizure cycle tracking can be applied to maximise the impact of seizure therapies and treatments.

Seizure cycle tracking opens up the opportunity to coordinate with other therapies to further improve people’s lives, potentially reducing the rate and severity of seizures.

Deep brain stimulation is a therapy where an implanted device constantly delivers electrical impulses to certain areas of the brain to reduce the frequency and severity of seizures. Guided by seizure cycle tracking, it could be possible to tailor the stimulation to only be delivered when a seizure is most likely to occur. By doing so, the severity of the event may be decreased or, even better, prevented altogether.

Certain medication can also be timed to be taken during periods of high seizure likelihood. Lifelong and constant anti-seizure medication come with many side effects. Drowsiness, not feeling like oneself, and nausea are commonly reported side effects.

Working with their primary physician, a patient could tailor when they need to take medication, minimising how long they experience side effects and allowing them to feel normal.

Reducing medication dependence provides alternatives to managing epilepsy, opening the doors to a higher quality of life.