A competition run on data science crowd-sourcing platform Kaggle has found the prediction of epileptic seizures is possible in far more people living with the condition than previously thought.
University of Melbourne researchers posted the competition on the platform in 2016, prompting data scientists from around the world to develop algorithms to predict seizures, based on data from a clinical trial in 2013.
Some 646 participants, grouped into 478 teams developed more than 10,000 algorithms aimed at distinguishing between the electrical brain activity observed between seizures and just prior to seizures.
The data was taken from trials of Seattle firm NeuroVista’s ‘seizure advisory system’. The system involves a series of electrodes implanted between the skull and the brain surface of a patient which monitor their electroencephalogram (EEG), essentially a measure of the brain’s electrical activity.
“We wanted to draw on the intelligence from the best international data scientists to achieve advances in epileptic seizure prediction performance for patients whose seizures were the hardest to predict,” said Dr Levin Kuhlmann, from the university’s Graeme Clarke Institute and St Vincent’s Hospital Melbourne.
The best algorithms from the competition have now been put to work on a much larger data set, which includes patients with a low level of seizure predictability, and found to be just as effective.
“These results indicate that clinically-relevant seizure prediction is possible in a wider range of patients than previously thought possible,” wrote Kuhlmann et al in their paper published yesterday in Brain: A Journal of Neurology.
Different algorithms contributed to Kaggle – which was founded in Melbourne and acquired by Google last year – in certain cases led to a 90 per cent improvement in seizure prediction performance, compared to previous results. This supports the concept of using different algorithms for different patients, Kuhlmann said.
“Epilepsy is highly different among individuals. Results showed different algorithms performed best for different patients, supporting the use of patient-specific algorithms and long-term monitoring,” he explained.
To that end, the researchers have now launched a website EpilepsyEcosystem.org to further refine and evaluate algorithms. It features a web-based data viewer and Python-based API to download it.
Once the ecosystem has determined the best algorithms, they will be entered into ‘the ultimate benchmark’ test, and run on a larger data set.
“Our results highlight the benefit of crowdsourcing an army of
algorithms that can be trained for each patient and the best algorithm chosen
for prospective, real-time seizure prediction. We
want people to find the best seizure prediction algorithms so that seizure
prediction can be made a reality for patients worldwide,” Kuhlmann
“It’s about bringing together the world’s best data scientists and pooling the greatest algorithms to advance epilepsy research.,” he said.
Less earthquake, more hurricane
The ability to accurately predict seizures holds huge promise for the more than 250,000 Australians living with epilepsy. Seizures and blackouts can affect consciousness, awareness and judgement, putting individuals at risk of accidents and injury.
Depending on the type of seizures and how often they occur, those with epilepsy may be forced to avoid driving, working with machinery, working around water and participating in sports.
There are also psychological effects.
“Living with constant uncertainty significantly contributes to increased anxiety in people with epilepsy and their families, never knowing when the next seizure may occur. Even people with well controlled epilepsy have expressed their constant concern, not knowing if or when they will experience a seizure at work, school, travelling or out with friends,” said Carol Ireland, chief executive of Epilepsy Action Australia.
“Any progress toward reliable seizure prediction will significantly impact the quality of life and freedom of choice for people living with epilepsy,” she added.
A number of Australian institutions are attempting to create seizure warning systems, and taking different approaches.
In May, University of Sydney researcher Dr Omid Kavehei from the Faculty of Engineering and IT said his team was “on track to develop an affordable, portable and non-surgical device that will give reliable prediction of seizures”.
Kavehi’s work is based around an algorithm which uses a convolutional neural network to generate optimised features for each patient.
Similar work is ongoing at the Deakin Software and Technology Innovation Laboratory (DSTIL) at Deakin University in collaboration with Royal Melbourne Hospital. They too are seeking to predict seizures from data collected by wearables.
“The hope is to make seizures less like earthquakes, which can strike without warning, and more like hurricanes, where you have enough advance warning to seek safety,” Kuhlmann added.