Toward Unified Biomarkers for Focal Epilepsy.

Area of Science:

• Neuroscience
• Computational Biology
• Medical Imaging

Background:

• Accurate localization of the epileptogenic network (EpiNet) is crucial for effective epilepsy treatment but is hindered by limited mechanistic understanding.
• Patient-specific EpiNets are shaped by complex pathologies, and combining biomarkers increases data dimensionality, risking overfitting and reducing interpretability.

Purpose of the Study:

• To hypothesize that core epileptogenic dynamics can be captured in a low-dimensional latent space derived from empirical data, independent of seizure recordings.
• To develop a simplified, interpretable probabilistic model for EpiNet localization.

Main Methods:

• Extracted 260 neuronal features from interictal stereo-EEG (SEEG) recordings in 64 epilepsy patients.
• Reduced feature dimensionality to 10 latent components using singular value decomposition.
• Developed a probabilistic EpiNet model requiring only two components, validated in independent patients.

Main Results:

• A classifier trained on 10 components simplified to a two-component probabilistic model, showing functional relevance (r²=0.5).
• The model captured time-varying epileptogenic dynamics in three independent patients during sleep-SEEG, with peak accuracies of 0.63, 0.85, and 0.94.
• Predictions were validated by tensor component analysis, demonstrating robustness across brain states and pathologies.

Conclusions:

• A robust low-dimensional representation of epileptogenicity exists, simplifying interpretation and biomarker integration.
• This approach offers a proof of concept for a unified framework for epilepsy biomarkers, enabling large-scale cohort analyses.
• The findings pave the way for improved epilepsy diagnosis and personalized treatment strategies.