PredictMed-epilepsy: A multi-agent based system for epilepsy detection and prediction in neuropediatrics

Area of Science:

• Neurology and Artificial Intelligence
• Pediatric Developmental Disorders
• Clinical Decision Support Systems

Background:

• Epileptic seizures frequently co-occur with developmental disabilities and cerebral palsy (CP).
• Early identification and management of epilepsy in these children are crucial for improved outcomes.
• Existing clinical support systems lack specialized modules for predicting epilepsy in CP populations.

Purpose of the Study:

• To develop and validate a deep-learning model, PredictMed-Epilepsy, for predicting epilepsy in children with CP.
• To identify key clinical factors associated with epilepsy in children diagnosed with developmental disabilities and CP.
• To integrate this predictive module into a broader multi-agent healthcare system for potential clinical decision support.

Main Methods:

• A longitudinal, multicenter, double-blinded descriptive study involving 102 children (12-18 years) with CP.
• Data collection included CP etiology, epilepsy type, spasticity, clinical history, communication, motor, and feeding abilities (2005-2021).
• The PredictMed machine-learning model was utilized to identify epilepsy predictors, adhering to TRIPOD guidelines.

Main Results:

• Factors significantly associated with epilepsy included CP etiology (prenatal > perinatal > postnatal), scoliosis, communication and feeding disorders, poor motor function, intellectual disabilities, and spasticity type (quadriplegia/triplegia > diplegia > hemiplegia).
• The PredictMed model demonstrated an average accuracy, sensitivity, and specificity of 82%.
• A computational phenotype for children with CP at risk of epilepsy was defined by the PredictMed model.

Conclusions:

• PredictMed-Epilepsy effectively identifies children with developmental disabilities and CP who are at higher risk for epilepsy.
• The model's findings highlight specific clinical characteristics that warrant closer monitoring for epilepsy.
• The developed Multi-Agent System (MAS) with the PredictMed-Epilepsy module, including real-time event processing, represents a novel advancement for patient monitoring.