Early Identification of Candidates for Epilepsy Surgery: A Multicenter, Machine Learning, Prospective Validation...

Area of Science:

• Neurology
• Artificial Intelligence in Medicine
• Clinical Informatics

Background:

• Epilepsy surgery is often delayed, impacting patient outcomes.
• Machine learning (ML) models were previously developed to expedite the identification of resective epilepsy surgery candidates.
• This study focuses on the prospective validation of these ML models.

Purpose of the Study:

• To prospectively validate machine learning models for early identification of epilepsy surgery candidates.
• To assess the accuracy and efficiency of ML algorithms in distinguishing patients who require resective epilepsy surgery.

Main Methods:

• A multicenter, prospective, longitudinal cohort study involving pediatric and adult epilepsy centers.
• Random forest ML models utilized diverse patient data including clinical notes, EEG/MRI reports, and healthcare utilization patterns.
• Model performance was evaluated using Area Under the Curve (AUC), positive predictive value (PPV), and negative predictive value (NPV).

Main Results:

• The ML models demonstrated high accuracy with an AUC of 0.91 in both pediatric and adult cohorts.
• The models identified potential surgical candidates significantly earlier than traditional evaluation timelines (median 2.1 years pre-surgery).
• Negative predictive value (NPV) was 1.00, indicating high confidence in identifying patients who do not require surgery.

Conclusions:

• Machine learning algorithms are effective in identifying candidates for resective epilepsy surgery earlier in the disease course.
• Despite specialized care, opportunities exist to further shorten the time to epilepsy surgery through advanced algorithms.