Watch all the Transform 2020 sessions on-demand here.
About 1.2 percent of people in the U.S. — and 3.4 million worldwide — have active epilepsy, and roughly one in 26 people will develop it in their lifetime. Not all suffer seizures the same — and for a third of patients, no medical treatment options exist. As for the remaining two thirds, the available treatments don’t always behave predictably, owing to the condition’s individualized nature.
Lack of measurement is a long-standing barrier to better outcomes. Studies show that one common source of data — written diaries — tends to be only 50 percent accurate. That’s why researchers at IBM are investigating an artificially intelligent (AI) system that could, using brain wave data in the form of electroencephalogram (EEG) readings, classify various types of seizures automatically.
It’s a relatively novel approach. Recent similar efforts include a competition sponsored by MathWorks, National Institutes of Health’s National Institute of Neurological Disorders and Stroke (NINDS), and American Epilepsy Society on Google’s Kaggle platform that challenged participants to train algorithms on seizure suffers’ EEG data. Participants in the competition and others working on the problem have looked at how to predict seizures, but comparatively few have attempted to distinguish among seizure types. The IBM scientists believe that faster and more accurate classification could hold the key to highly tailored treatment and disease management.
“Accurate classification of seizure types plays a crucial role in the treatment and disease management of epileptic patients,” the researchers wrote in a paper published on the preprint server Arxiv.org (“Machine Learning for Seizure Type Classification: Setting the benchmark“). “This technology could support automatic seizure type logging in digital seizure diaries. Such seizure diaries could then be used to improve the performance of clinical trials through more efficient and reliable patient monitoring for endpoint detection, adherence control, and patient retention.”
June 5th: The AI Audit in NYC
Join us next week in NYC to engage with top executive leaders, delving into strategies for auditing AI models to ensure fairness, optimal performance, and ethical compliance across diverse organizations. Secure your attendance for this exclusive invite-only event.
The paper’s authors sourced Temple University’s publicly available TUH EEG Seizure Corpus — a dataset containing the occurrence and types of 2,012 seizures across eight categories — to “teach” the AI system to recognize seizures, and to test its precision. (They ignored myoclonic seizures because of their “very low count,” which left them with seven classes of seizures in total.) About 60 percent and 20 percent of the seizure data was divided into training and validation sets, respectively, with the remaining 20 percent serving as a test set.
The team evaluated a variety of machine learning classifiers but found that one — k-NN, the k-nearest neighbors algorithm — outperformed the others. It achieved an F score (which measures accuracy) of 0.907, meaning it managed to accurately determine seizure type about nine out of 10 times.
“[W]e hope that automatic classification of seizure types will improve long-term patient care, enabling timely drug adjustments and remote monitoring,” the researchers wrote. “To the best of our knowledge, this is the first work showing that machine learning techniques can be successfully used for automated multi-class seizure type classification.”
