Australasian Science: Australia's authority on science since 1938

Cloudy with a Chance of Seizure

Credit: Rainer Fuhrmann/Adobe

Credit: Rainer Fuhrmann/Adobe

By Katrina Dell

Just as we check the weather forecast to plan our daily activities, people with epilepsy will soon be able to check personalised seizure forecasts to determine their risk and take necessary precautions.

Imagine waking up each day and wondering if you will have a seizure, knowing it could strike you down at any time or place. Maybe you will black out while you are swimming, crossing a busy street or holding a small child.

This frightening uncertainty is the reality for many people living with epilepsy. A device or app that could provide a real-time indication of the likelihood of a seizure would not only allow people to take precautionary measures to avoid injury but it could also be used to automate treatments and therapies. Led by Prof Mark Cook, a team of researchers at the University of Melbourne are developing tools for seizure forecasting to help improve the quality of life of people living with epilepsy.

Epilepsy is a neurological disorder characterised by recurrent seizures. It affects approximately 50 million people worldwide.

There are millions of neurons in the brain that communicate via electrical impulses to convey messages, regulate body systems and produce movements, thoughts and feelings. A seizure is essentially the manifestation of an electrical glitch in the brain. They occur when normal patterns of activity are disrupted by abnormal and excessive neuronal activity.

The way a seizure presents can vary greatly depending on the affected regions of the brain. For example, a person having a seizure could convulse and lose consciousness, or instead remain conscious but have difficulty speaking.

About one-third of people diagnosed with epilepsy cannot control their seizures with the medications available, and those that are controlled are at risk of breakthrough seizures. Sadly, the prevalence of seizure control has not improved in several decades despite the introduction of many new anti-epileptic drugs. It’s clear that we need to adopt a different approach to improve seizure management.

The unpredictability of seizures is one of the greatest burdens for people living with epilepsy, so the development of seizure forecasting for both seizure warnings and automated therapeutic devices is of great interest. Three major components are required to make seizure forecasting a reality.

First, we need to identify patterns related to seizure timing that can be used to create a reliable seizure forecasting algorithm. Second, we need to create a device that can collect the information needed for the algorithm to provide an estimate of seizure likelihood. Finally, we need a device or app that can relay the seizure likelihood estimates to the relevant people. Our team has made significant headway on all three components.

Studies indicate that the brain enters a state of increased excitability prior to a seizure. This suggests that some features of brain activity can be used to predict when a seizure is likely to occur.

However, previous attempts at seizure forecasting have mostly failed. This is likely due to the short duration of monitoring that is typically used in a hospital setting. People with epilepsy undergo routine brain monitoring via scalp electroencephalographs, where approximately 21 electrodes are placed on the scalp to record brain activity for periods of up to 7 days. These short-duration recordings may capture only a handful of seizures per patient, which is not sufficient to identify patterns that occur in brain activity leading up to seizures.

To compare across more seizures, researchers often pool data from multiple patients. However, this is problematic as the triggers and mechanisms that lead to seizures are highly variable between patients, so it is likely that the patterns of brain activity that precede seizures will also vary. Long-duration recordings that capture a large number of seizures per patient are necessary for identifying patterns that can be used to forecast seizures.

A landmark study carried out in Melbourne captured the world’s longest continuous brain recordings in 15 patients with drug-resistant epilepsy. The data were recorded over a 3-year period using electrodes surgically implanted to sit on the surface of the brain. The trial demonstrated for the first time that seizure forecasting is possible, though it was not successful for all patients. Continued work by our team using the same dataset has since enhanced seizure-forecasting performance, making it viable for more patients.

Our results show that existing algorithms of seizure forecasting can be improved by incorporating factors such as sleep and weather, along with patterns in seizure occurrence and various aspects of brain activity. Importantly, these patterns are highly patient-specific, and could not have been observed with standard short-term recordings.

Our team has recently identified rhythms in brain activity that are closely linked to seizure occurrence. This is exciting as the rhythms help us to track the brain’s susceptibility to a seizure more accurately. We therefore anticipate that the incorporation of these new findings with our existing algorithms will significantly improve seizure-forecasting capabilities.

To collect the information needed for seizure forecasting, our team has recently designed a new implantable device for long-term brain monitoring and a seizure-forecasting app. The brain-recording device sits behind the ear, with three electrodes running under the scalp and an external battery worn behind the ear. Brain signals will be recorded continuously and then sent to a smartphone app or computer capable of real-time seizure forecasting.

The app we have designed will act as a digital seizure diary where patients can log information about their seizures as well as other relevant lifestyle factors, such as sleep, stress, exercise and alcohol intake. The information will then be combined with local weather and other external data to create a powerful and personalised seizure-forecasting tool.

Soon our app users will be able to check their daily risk of seizure just as they check the weather forecast for the chance of rain. It is our hope that people with epilepsy will be able to use the app to minimise their risk of seizure by planning daily activities accordingly, such as avoiding alcohol, exercise or other seizure triggers on the days when their seizure likelihood is high. The app could also be used to generate warnings when there is a high-risk of seizure.

Depending on personal preferences, the phone could vibrate or produce a tone when the seizure likelihood reaches a selected point, enabling the user to take additional safety precautions when needed. In this way, a reliable seizure forecast could help people with epilepsy to regain their independence.

The potential benefits of seizure forecasting are clear, but when can you expect to see our seizure-forecasting app become publicly available? Already our team has released the Beagle Health Tracker app, which acts as a digital seizure diary. By the end of the year we hope to incorporate a basic forecasting function that uses long-term patterns from seizure occurrence, behavioural, environmental and physiological data collected by the app.

Once long-term brain monitoring devices are in use, we can start to integrate trends in brain activity to refine forecasts. Ultimately, we hope to use seizure forecasting to regulate treatments and therapies so that people with epilepsy can control their seizures without even thinking about it. The technology could be life-changing for people living with epilepsy.


Katrina Dell is a Research Fellow in the Department of Medicine, St Vincent’s Hospital at the University of Melbourne.