A Comparison of Two Deep Learning Approaches to Distinguish Functional Dissociative from Epileptic Seizures Using...

Area of Science:

• Neurology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Differentiating motor epileptic seizures (ES) from functional dissociative seizures (FDS) is a diagnostic challenge.
• Video electroencephalography (vEEG) is the gold standard but has limitations in availability and interpretation.
• There is a need for objective, automated tools to aid in seizure diagnosis.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for differentiating motor ES from FDS using only video data.
• To compare the performance of a pose-estimation ML model with a 3D convolutional neural network (CNN) model.

Main Methods:

• Retrospective study of 106 seizure event videos from 10 patients.
• Development of two ML models: one using pose-estimation and another using an end-to-end 3D CNN.
• Performance evaluation using area under the receiver-operating characteristic (AUROC) and precision-recall (AUPRC) curves, sensitivity, precision, and accuracy.

Main Results:

• Both ML models distinguished between ES and FDS better than chance.
• The CNN model achieved higher performance: AUROC 0.78, AUPRC 0.84, sensitivity 0.82, precision 0.82, and accuracy 0.80.
• The pose-estimation model achieved an AUROC of 0.71, AUPRC 0.53, sensitivity 0.90, precision 0.50, and accuracy 0.62.

Conclusions:

• 3D CNN models demonstrate superior performance over pose-estimation models for differentiating motor ES and FDS from video.
• These ML tools show potential as adjunct diagnostic aids for rapid, objective seizure evaluations.
• Further studies are needed, but these models could reduce reliance on immediate neurologist interpretation.