June 8, 2022 | Novel network analysis technology that uses only 10 minutes of resting state electrophysiological recordings has been shown to localize seizure onset brain regions and predict seizure outcomes in epilepsy patients. The breakthrough emerges from a collaboration between researchers at Carnegie Mellon University, the University of Pittsburgh Medical Center, and Harvard Medical School, and is detailed in an article that published last month in Advanced Science (DOI: 10.1002/advs.202200887).
The new technique could shave as much as 10 days off the time it takes to complete an invasive procedure required of individuals who will be treated with either neuromodulation or ablation therapy, or surgery to remove the seizure-originating tissues, says lead researcher Bin He, professor of biomedical engineering at Carnegie Mellon. This would make hospital stays considerably shorter, if not eliminate them altogether, but in either case there would also be a marked reduction in associated medical costs.
Although the method has so far only been tested in a small group of hospitalized patients, in the future recordings could perhaps be taken in an outpatient setting, he says. Physicians could thereby make the decision right away whether to operate.
The need for stereotactic-electroencephalography (SEEG), which involves drilling into the skull to place recording electrodes atop the brain, remains the gold standard here, He emphasizes. But instead of waiting for seizures to come during the intracranial monitoring process, the new method estimates the seizure onset zone (SOZ) and predicts prognosis outcome from an analysis of brief, resting-state data segments.
Information flow gets extracted across the recording electrodes and predictions are based on the different levels of functional connectivity between brain tissue, he explains. As He and his colleagues have previously shown (Annals of Neurology, DOI: 10.1002/ana.25583), the notably greater difference in information flow from non-seizure-generating tissue to seizure-originating tissue is much larger than the inverse direction and often leads to a seizure-free surgical outcome.
The team’s one fundamental mechanistic hypothesis is that during the resting state, when a seizure is not occurring, the non-SOZ is suppressing the seizure-generating region, he continues. But once a seizure occurs, the push-pull dynamics flip, and information flow moves in the other direction.
“The mechanism for generating seizures is very complex, and not something everyone agrees on,” says He. The theory on which their analysis technique is based runs counter to conventional thinking holding that seizures are like a spot that gets hyperexcited and gradually spreads to other areas.
The method estimates the information flow between every single pair of electrodes, as inferred from directional connectivity, He says. In a cohort of 27 drug-resistant focal epilepsy patients, the newly published retrospective study suggests brief resting-state SEEG data has the potential to facilitate the identification of the SOZ and may eventually predict seizure outcomes without the need for long-term recordings of in-progress seizures.
A balanced random forest model was integrated with resting-state connectivity to localize the SOZ and predict seizure outcomes with, respectively, 88% and 92% accuracy.
Rather than treating a focal spot, as has traditionally been done, He and his team believe that any therapeutic approach—be it drugs, resection surgery, laser ablation, or neuromodulation (injection of electromagnetic or acoustic energy)—needs to take network behavior into consideration.
He says he next hopes to team up with neurosurgeons and neurologists to conduct a prospective clinical study enrolling drug-resistant epilepsy patients prior to surgery to confirm that information from resting-state SEEG optimizes their treatment.
If brought into clinical use, the technique would be designated “software as a medical device” by the U.S. Food and Drug Administration, says He. Preferentially, the technology will be licensed to an interested company, but otherwise the plan is to independently bring the customized software and data analysis techniques to market.
It would be a welcome addition to the medical arsenal. Epilepsy affects about 70 million people around the world and more than 3.4 million Americans. Of those affected, roughly a third cannot be treated by drugs alone, he says.
The research to date serves to highlight the potential of the broader field of neuroengineering to help solve clinical neuroscience and neurology problems, He notes. It’s a relatively new discipline that applies physical science and engineering methods to solving problems in neuroscience and clinical brain science, as illustrated in theoretical network analysis of electrophysiological signals in brain circuitry.