How EEG and machine learning are transforming epilepsy clinical trials

Epilepsy is a brain disorder that triggers recurring seizures. It is the fourth the most common neurological disorders in the world, according to the Epilepsy Foundation. The Centers for Disease Control and Prevention estimates that 65 million people worldwide have active epilepsy. In 2015, 1.2% of the total U.S. population — 3 million adults and nearly 500,000 children — had active epilepsy.

There are many different causes of epilepsy, including genetics, head trauma, brain abnormalities, infection, prenatal injury and developmental disorders, such as autism.

Seizure symptoms vary greatly and can manifest in a person as uncontrollable limb movements, staring, muscle stiffness, confusion and loss of consciousness or awareness.

These symptoms are not mutually exclusive. Some patients with epilepsy experience multiple types of seizures.

Epilepsy can impact a person’s physical safety, relationships, work, independence and other aspects of their lives. While there is no known cure for epilepsy, there are various treatments for this neurological disorder, including seizure medications, seizure devices and surgery. Unfortunately, treatment doesn’t work for every person with epilepsy.

Recent breakthroughs in neurodiagnostics and digital biomarkers provide researchers with unprecedented insights into brain activity. This data is expected to pave the way for novel therapies and individualized treatments for epilepsy.

EEG data at scale

EEG (electroencephalography) is currently considered the gold standard clinical diagnostic for epilepsy-related disorders. However, technology limitations have historically limited researchers’ ability to collect robust sets of EEG data. This lack of EEG data and derived endpoints has severely restricted research. It has limited the study of epilepsy causes and therapies.

Now, quantitative tools can measure disease burden in ways missed by patient or caregiver seizure diaries and other subjective measures of disease. Applying scalable machine learning tools to EEG data allows for the analysis of more brain activity than previously thought possible.

Collecting vast amounts of data points from epilepsy patients helps researchers. They can gain insights into individual causes and understand different epilepsy syndromes better. This may lead to the development of individualized treatments and therapies.

Further, the ability to analyze EEG data at scale will accelerate clinical trials, allowing beneficial drugs to reach patients with epilepsy faster. Data can determine patient safety for trial enrollment and if they have specific disease features. It can also show if treatment modulates those endpoints.

Large EEG data sets unveil epilepsy insights

Collecting EEG endpoints automatically and at scale provides data that might not otherwise be available. There are people with epilepsy who may have seizures in their sleep and not remember them when they wake up. And children may have seizures when their parents aren’t with them. Quantitative endpoints provide enough data to advance research into epilepsy causes and treatments.

Large data sets allow researchers to understand the linkages between genetic epilepsies and variables regarding patients’ motor skills, cognitive skills, language capabilities and other abilities. EEG is helping us understand molecular genetic change on an individual level and how neural circuits evolve with a particular mutation. Quantitative EEG data will help bridge the gap between the genetic and molecular levels and the subjective, visible patient-reported outcomes level.

EEG endpoints will increase our understanding of different epilepsy syndromes by capturing changes and different patterns in brain data. For example, EEG data could help clinicians determine if there is underlying epileptiform activity in children displaying abnormal behaviors, and thereby inform decision-making on further genetic testing.

Accelerating clinical trials

Applying artificial intelligence and machine learning to EEG data can accelerate clinical trials and identify particular therapies that might be best for individual patients with a specific form of epilepsy. Clinical trials for epilepsy treatments have historically relied on patient seizure diaries. Not only can these diaries be “noisy,” it takes a lot of time before they have sufficient statistical weight. Quantitative EEG endpoints can enable sponsors to more quickly identify patients with specific EEG features for participation in a trial. Once that trial is under way, EEG data can measure the efficacy of the drug being developed in a far shorter period of time.

EEG data already shows a link between genetics, the types of mutation and patient outcomes. It explains the variability we see and the heterogeneity in their outcomes.

We’re finding more and more in developmental and epileptic encephalopathies, these rare epilepsies, that EEG correlates with poorer or less severe outcomes. That’s incredibly important information for clinical trials to access and use to target one of those two populations specifically.

Conclusion

We now have the technology to capture EEG data inexpensively and at scale to enable earlier diagnoses. This will be invaluable for children who don’t have visible seizures but start missing motor milestones and show slightly abnormal behaviors. An EEG can determine whether the child has epilepsy. Quantitative EEG data that captures the pathology of epilepsy and whether it is being modulated by a treatment can then serve as endpoints in clinical trials. Someday, that may result in the approval of a drug based on the way it changes brain activity on a really short time scale.

Jacob Donoghue, MD, Ph.D., is the CEO and co-founder of Beacon Biosignals.