Developing a machine learning model to assist in predicting treatment success in children with drug-resistant...

Area of Science:

• Neurology
• Artificial Intelligence
• Pediatric Medicine

Background:

• Drug-resistant epilepsy (DRE) in children presents significant treatment challenges due to patient variability.
• Current antiepileptic drug (AED) selection often struggles to achieve consistent seizure reduction in pediatric DRE.
• Personalized treatment strategies are crucial for improving outcomes in pediatric epilepsy.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting treatment success in pediatric patients with DRE.
• To identify the optimal machine learning algorithm for forecasting treatment outcomes.
• To create a decision support tool for clinicians managing pediatric DRE.

Main Methods:

• An ambispective cohort of 215 pediatric patients with DRE was studied.
• Data included clinical information, electroencephalography (EEG), and Magnetic Resonance Imaging (MRI).
• Machine learning algorithms (SVM, DT, RF, GB) were employed and compared for predictive performance.

Main Results:

• Machine learning models integrated diverse data types to predict treatment success.
• Comparative analysis identified the most effective algorithm for forecasting seizure control.
• The study establishes a novel approach for predicting therapeutic response in pediatric DRE.

Conclusions:

• Machine learning offers a promising approach to personalize AED selection for pediatric DRE.
• The developed model can aid neurologists in predicting seizure control and guiding therapeutic adjustments.
• This research pioneers the use of integrated data with machine learning for DRE management in Indonesia.