Photo Credit: Mr. Suphachai Praserdumrongchai
According to Benjamin D. Wissel, MD, PhD, and colleagues, machine learning algorithms effectively identify candidates for resective epilepsy surgery earlier in the disease course, potentially shortening the time to surgery. The researchers aimed to validate the use of machine learning models. The prospective validation involved a multicenter cohort study, utilizing random forest models trained on various patient data, including neurology notes, EEG and MRI reports, visit patterns, and medical orders. The study included both pediatric and adult epilepsy centers, with promising results. The machine learning algorithms demonstrated high accuracy, with an area under the curve score of 0.91 for pediatric and adult cohorts. Positive predictive values were low, indicating cautious interpretation of high scores, while negative predictive values were excellent. Importantly, the models identified surgical candidates significantly earlier than traditional evaluation methods, potentially shortening the time to surgery by several years. This study presents Class II evidence supporting the efficacy of machine learning algorithms in distinguishing epilepsy patients requiring surgery, highlighting their potential for improving patient care by reducing delays in surgical intervention.
