An Artificial Neural Network Prediction Model for Posttraumatic Epilepsy: Retrospective Cohort Study

Area of Science:

• Neurology
• Artificial Intelligence in Medicine
• Biostatistics

Background:

• Posttraumatic epilepsy (PTE) is a frequent complication following traumatic brain injury (TBI).
• Accurate identification of high-risk patients is crucial for effective PTE management.
• Existing artificial neural network (ANN) prediction models for PTE are limited.

Purpose of the Study:

• To develop and validate an ANN model for predicting PTE risk in TBI patients.
• To compare the performance of the ANN model against traditional prediction methods.

Main Methods:

• A 5-fold cross-validation approach was used to train and test the ANN model on a TBI patient cohort.
• 21 independent variables were utilized as input neurons.
• The model's performance was evaluated using sensitivity, specificity, accuracy, and area under the ROC curve, with external validation on two independent cohorts.

Main Results:

• The ANN model demonstrated strong predictive capabilities with an area under the ROC curve ranging from 0.859 to 0.907 across cohorts.
• Accuracy varied across cohorts, with the highest in the training set (0.557).
• The ANN model showed significantly higher accuracy compared to a traditional nomogram model (P=.01).

Conclusions:

• The developed ANN model effectively predicts PTE risk in TBI patients, outperforming traditional statistical models.
• While promising, the model's calibration requires improvement through validation on larger datasets.
• Further refinement of the ANN model is recommended for enhanced clinical utility in PTE risk assessment.