Using Artificial Intelligence Algorithms to Predict Seizure Control Following Brain Surgery: Lara Jehi, MD, MHCDS

WATCH TIME: 6 minutes

"The attractive thing with scalp EEG is that this is the bread and butter of evaluating any patient with epilepsy, anywhere. So every patient who is getting epilepsy surgery is going through that procedure. If we can learn anything from a scalp EEG, whatever that knowledge is, it will be generalizable and helpful to the population at large. That’s what drove this study."

For years, surgical brain resection has been a promising approach for patients with drug-resistant epilepsy (DRE); however, only about half who undergo a procedure achieve sustained seizure freedom. Identifying those who are likely to experience recurrent seizures after resection has been a challenge thus far for clinicians and researchers alike. Previously conducted preclinical data from members at the Cleveland Clinic showed that the window of time immediately before and after a seizure (“peri-ictal”) represents a unique brain state with implications for clinical outcome prediction.

A follow-up study, published recently in Nature, highlighted the predictability of a machine learning model-building experiment that used 5 minutes of a peri-ictal scalp EEG data. The study, which included 294 surgical patients who underwent temporal lobe resection for seizures, showed that this approach was successful in making accurate predictions of postoperative seizure outcomes, demonstrated by area under the operating receiver curve of 0.98. In the study, a decision curve analysis also showed that compared with the prevalent clinical-variable based nomogram, use of the EEG-augmented approach could decrease the rate of unsuccessful brain resections by 20%.

Following the publication, study investigator Lara Jehi, MD, MHCDS, sat down with NeurologyLive^® to provide more background on the machine learning model and the unique aspects of the study design. Jehi, an epilepsy specialist and Cleveland Clinic’s Chief Research and Information Officer, emphasized the importance of a large, well-characterized patient cohort with consistent follow-up and the choice of scalp EEG– commonly used, non-invasive test in epilepsy care–as the data source. In addition, she provided background on some of the major differences between unsupervised and supervised machine learning, as well as the use of AutoML to streamline the study, enabling efficient identification of the top-performing algorithms and enhancing the model’s predictive accuracy.

REFERENCE
1. Sheikh SR, McKee ZA, Ghosn S, et al. Machine learning algorithm for predicting seizure control after temporal lobe resection using peri-ictal. Nature. 2024(14):21771. doi:10.1038/s41598-024-72249-7