AI-Driven Neurodiagnostics: A Scalable Framework for EEG Anomaly Detection Using a Distributed-Delay Neural Mass...

Area of Science:

• Computational Neuroscience
• Artificial Intelligence
• Clinical Neurodiagnostics

Background:

• Limited pathological electroencephalogram (EEG) datasets pose challenges for clinical neurodiagnostics.
• Integrating biophysically grounded neural simulations with AI offers a novel solution.

Purpose of the Study:

• To develop an AI-driven framework for generating synthetic EEG signals.
• To enhance anomaly detection in EEG data for improved neurodiagnostics.

Main Methods:

• Utilized a Distributed-Delay Neural Mass Model (DD-NMM) to simulate healthy and pathological EEG.
• Employed systematic parameter tuning and data augmentation for signal enrichment.
• Integrated supervised classification and unsupervised one-class anomaly detection.

Main Results:

• Achieved over 95% accuracy on synthetic EEG data.
• Demonstrated over 89% accuracy on empirical EEG data from epilepsy patients and healthy volunteers.
• Successfully replicated healthy and pathological brain states.

Conclusions:

• The AI framework effectively bridges computational neuroscience and AI for EEG anomaly detection.
• This approach advances early seizure detection, adaptive neurofeedback, and brain-computer interfaces.
• Theory-driven simulation combined with machine learning addresses critical gaps in medical AI and clinical neuroengineering.