New AI technique can identify seizure types, including rare forms of epilepsy

More than 3.4 million people in the United States and 65 million people worldwide have epilepsy, a neurological disorder that affects the nervous system and causes seizures. One in 26 people will develop epilepsy at some point in their lives, and one in 1,000 people with epilepsy die unexpected deaths each year.

As with many pathologies, treatment of epilepsy begins with early detection. The World Health Organization estimates that 70% of people with epilepsy could live seizure-free if properly diagnosed and treated.

Over the years, machine learning techniques have been developed to detect and classify seizures from electroencephalography (EEG) signals captured using electrodes on the brain, looking for correlations too complex to that humans can face it alone.

However, these systems have difficulty detecting rare forms of epileptic seizures. This is because AI relies on data to learn patterns and make predictions: insufficient examples of these rarer crises limit its ability to function properly in less common cases.

Now, USC researchers have developed an AI system to identify epilepsy by analyzing brain interactions, improving the diagnosis of rare and complex cases. The system, presented at the Advances in Knowledge Discovery and Data Mining (PAKDD) conference in May 2024, and published on the preprint server arXivdemonstrated a 12% improvement over state-of-the-art models.

By integrating multiple sources of information typically overlooked by AI systems in epilepsy detection, including the position of EEG electrodes and the brain regions they monitor, AI can identify patterns or features that indicate when a crisis is likely to occur.

This technique also helps the system generate accurate results with less data, even in rare seizure types for which there may only be a few examples in the training data.

“Usually for the simplest use cases, an AI system can tell if someone has had a seizure since it's a simple binary classification,” said co-author Cyrus Shahabi , professor of computer science, electrical engineering and space sciences. “But there are different types of seizures, rarer, which are not easy to classify: existing techniques have low precision in this task.”

Take, for example, atonic seizures, a rare type of seizure that often affects children and triggers a sudden loss of muscle control and collapse. In this case, the system would examine spatial relationships across brain regions and prioritize brain areas involved in muscle control, such as the motor cortex, basal ganglia, cerebellum and brainstem, to identify patterns. of activity indicating atonic seizures.

“In our framework, we have the spatial relationships, semantics and descriptions of each part of the brain,” said lead author Arash Hajisafi, a computer science doctoral student supervised by Shahabi. “All of this information is extracted to help the model understand the relevant characteristics of this type of seizure. So even if you feed the neural network a small amount of samples, it will still learn.”

The objective is not to replace doctors, the researchers specify, but to supplement their knowledge in cases that are difficult to detect. For Paul Thompson, a neuroscientist at USC and professor of neurology who was not involved in the study, this is a welcome development that could be a “game changer” in clinical neurology.

“Understanding seizure types is crucial for early treatment, but recordings of brain activity are extremely complex,” Thompson said. “This breakthrough brings the power of AI to detect patterns that a human would struggle to identify, making this task easier, faster and more reliable for clinicians.”

One day, researchers hope this technology will be integrated into wearable sensors that transmit information to a smartphone.

“Brain seizures occur very suddenly and detecting them earlier could really save lives,” Shahabi said. “The system could trigger an alert if it detects irregularities in brain waves. This would open up incredible opportunities for the diagnosis and treatment of epilepsy.”