Skip to Main Content


Research Applies Machine Learning/AI to Predict Consciousness for Driving in People with Epilepsy

September 20, 2022
by Allison Greco

The road to obtaining a driver’s license is often a linear one punctuated by critical, mandatory assessments: pass the knowledge, vision, and road test, and then – literally and figuratively – drive off into the sunset.

Upon fulfilling the necessary state requirements, operating a motor vehicle can seem like a part of everyday life for people with average neurological function. For people with epilepsy, though, getting behind the wheel presents its own set of unique obstacles.

Many people with epilepsy have more severe and obvious seizures that would clearly preclude driving. Others, though, may experience milder symptoms with time or medication, and therefore hope that they can one day safely drive. However, brainwave EEG tests may still indicate brief episodes of abnormal brain activity such as “spike-wave discharges” (SWDs). While they are not always problematic, SWDs can impair consciousness, but to what degree?

A new study published in September 2022 in Annals of Clinical and Translational Neurology may guide doctors in how to proceed when they observe SWDs on EEG, but are unclear of the patient’s neurological capacity for driving. Led by Principal Investigator Hal Blumenfeld, MD, PhD, researchers have harnessed the power of machine learning/artificial intelligence (AI) to address the shortcomings of traditional testing.

“Specialized behavioral testing can determine whether or not patients can respond normally during SWDs,” said Dr. Blumenfeld. “However, this kind of split-second behavioral testing during SWDs can usually only be done in a research setting. Meanwhile, regular brainwave EEG measurements without behavioral testing is very widely available.”

With brainwave EEG as an available resource, Dr. Blumenfeld’s team applied machine learning/AI – as opposed to behavioral testing – to predict whether people have impaired consciousness during SWDs.

Having logged 350 SWDs in 75 epilepsy patients over an eight-year period, researchers used the data to teach the machine learning/AI how to tell the difference between EEG with SWDs that do or do not cause impaired consciousness.

“With this approach, our AI classifier…was 100% correct when it classified an EEG as ‘safe.’ On the other hand, the AI only predicted that it was safe for people to drive 35% of the time when their SWDs caused no impairment on behavioral observations. So, 65% of people who may be able to drive based on behavior would not be able to drive based on the AI using EEG alone,” said Dr. Blumenfeld.

The results from this study prove that EEG, even without behavioral testing, can accurately predict consciousness for driving. It paves the way for more comprehensive future studies, where larger datasets could fine tune the AI classifier.

Although more data is needed to better account for safety, the possibility of a better quality of life and freedom of movement might not be that far down the road for people with epilepsy.

Submitted by Allison Greco on September 20, 2022