An Integrated Multi-Scale Computational Framework for the Precise Localization of Infection-Induced Epileptic Foci

This paper introduces a multi-scale computational framework for the precise localization of infection-induced epileptic foci. Integrating neurobiology, biochemistry, neuroengineering, and pharmacology, it fuses hybrid CNN-LSTM models for EEG analysis and U-Net for MRI segmentation with mechanistic models based on reaction-diffusion kinetics and wave propagation. A Bayesian inference engine enables probabilistic estimation of focal parameters from multi-modal data, supported by explicit posterior and likelihood formulations. The framework includes a pharmacological sub-model for therapeutic simulation and is validated through an *in silico* proof-of-concept with fully reproducible Python code. Sensitivity analysis using Sobol indices quantifies parameter impacts, while preprocessing, robustness testing, and regulatory compliance ensure clinical translatability. Achieving spatial localization error of 1.2 mm and AUC of 0.96, the framework significantly outperforms existing methods and paves the way for patient-specific interventions in infection-related epileptogenesis, advancing precision neurology.