Predicting cognitive impairment in outpatients with epilepsy using machine learning techniques.

Area of Science:

• Neurology
• Artificial Intelligence in Medicine
• Cognitive Neuroscience

Background:

• Cognitive dysfunction is common in epilepsy, yet prediction models are scarce.
• Assessing cognitive function in epilepsy patients is crucial for management and quality of life.

Purpose of the Study:

• To establish an efficient workflow for predicting cognitive assessment outcomes in epilepsy outpatients.
• To identify key clinical markers for predicting Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA) scores.

Main Methods:

• Utilized data from 441 epilepsy outpatients, with 433 meeting criteria for analysis.
• Employed a combination of Random Forest (RF) and Redundancy Analysis (RDA) algorithms for feature selection and model optimization.
• Cross-validation and resampling techniques were applied to refine the predictive model.

Main Results:

• The Random Forest algorithm demonstrated superior performance in predicting both MMSE and MoCA outcomes.
• Seven key features were identified: sex, age, age of onset, seizure frequency, brain MRI abnormalities, EEG epileptiform discharge, and drug usage.
• This optimal feature combination showed high efficiency in predicting cognitive assessment outcomes.

Conclusions:

• A robust predictive model for cognitive function in epilepsy patients has been developed.
• The identified clinical markers provide valuable insights for personalized patient management and prognosis.
• This workflow enhances the ability to predict and potentially mitigate cognitive decline in individuals with epilepsy.