Reconstructing the Path of a Seizure

Epilepsy is one of the most common neurological diseases in the world, afflicting more than 50 million people. While most forms can be treated with therapeutics, patients with drug-resistant epilepsy often require surgery to treat their seizures.

In these instances, neurosurgeons typically cut out the portion of the brain responsible for triggering seizures. But even these invasive methods are only effective about 60 percent of the time.

A new tool developed by neural engineers at Duke University could help improve these surgical outcomes by enabling neurosurgeons to more accurately pinpoint the areas in the brain where the seizures originate.

The research appeared December 3 in the journal Brain.

To identify the brain region responsible for causing a patient’s epilepsy, physicians first use an EEG to record brain activity during a seizure by attaching electrodes around the patient’s scalp. If this does not provide enough information, they use a stereo-EEG, which requires the insertion of electrodes into suspected brain regions via small holes drilled into the skull. While more invasive, the stereo-EEG provides a higher resolution recording of brain activity during a seizure than does EEG.

“If the brain is Europe, the stereo-EEG can tell us if the seizure is happing in Brussels or Berlin. If the signal is coming from Brussels, then the surgeon removes Brussels,” said Warren Grill, the Edmund T. Pratt, Jr. School Distinguished Professor of Biomedical Engineering.

But these stereo-EEG recordings don’t show the full path of the seizure as it spreads through the brain; they only show the area with the highest activity. While the area of with the largest signals may be indicative of the seizure-causing region, it can also be the result of signals combining as the seizure propagates through neural pathways and not indicative of the region that needs to be removed.

“Say the electrode that gives you the biggest signals happens to be in Brussels, but the activity actually started in Bruges,” said Grill. “So, when you remove Brussels, the patient doesn’t get better because you didn’t cut out the region causing the problem.”

To resolve this, Brandon Thio, a recent PhD graduate from the Grill lab and first author of the paper, developed TEDIE. Short for Temporally Dependent Iterative Expansion, the algorithm collects signals from every implanted electrode and reconstructs how neural activity travels and changes during a seizure.

“All you need is imaging data from the patient’s brain and recordings from the stereo-EEG to allow the algorithm to create a movie showing where the seizure originated and how it moved through their brain,” said Thio. “We don’t just tell you where it starts—we also tell you how big it is. If you’re going to go through invasive brain surgery to cut out part of your brain, you want to make sure you get it all out on the first try.”

Once developed, the team tested TEDIE on simulated seizure with known sizes and locations and demonstrated performance that substantially outpaced other current algorithms. Then they applied TEDIE to patient data by analyzing the stereo-EEG recordings from 46 epilepsy patients from Duke University Medical Center and the University of Pennsylvania. This analysis showed that TEDIE’s reconstructions accurately identified regions that were removed in patients who showed no epilepsy symptoms after surgery.

The algorithm also enabled identification of location of seizure origin that were different that the regions that were removed in patients who continued to show symptoms after surgery. Grill, Thio and their collaborators used TEDIE to identify potential new surgical targets in 12 out of 23 patients with epilepsy symptoms that persisted after surgery.

Buoyed by this success, the team hopes to expand their work by introducing TEDIE into clinical studies. They also want adapt the algorithm to be used for conventional EEG and thereby reduce the need for the invasive stereo-EEG and provide a more accessible tool for non-expert epilepsy centers. And, the team adds, TEDIE is more than capable of mapping brain activity to support basic neuroscience studies.

“Epilepsy is a very complex disorder. In some individuals, physicians will remove parts of the brain and the patient will get better, but then a year later seizures will return,” said Thio. “TEDIE likely won’t bring the efficacy up to 100 percent, but we hope that it improves on the current 60 percent clinical success rate.”

The team is grateful for the funding and support for this work provided by Duke MEDx, a Duke CTSA grant (UL1 TR002553), and an NIH F31 grant (NS124094).