Vagus-Decipher AI: Neural Decoding of Vagus Nerve Electrophysiology for Real-Time Prediction of Systemic Inflammatory Storms

Vagus-Decipher AI is a physics-informed neural decoding framework engineered to extract latent immunological state estimates from raw cervical vagus nerve electroneurogram (ENG) recordings for real-time prediction of systemic inflammatory storms. The framework integrates three mathematically grounded components: (1) the Adaptive Wavelet Isolation Engine (AWIE), a multi-resolution signal decomposition pipeline isolating immune-afferent spike trains from dominant cardiorespiratory noise at −20 to −35 dB SNR; (2) the Neuro-Immune State-Space Decoder (NISSD), a physics-constrained recurrent neural operator mapping inhomogeneous Poisson firing dynamics λ(t) into a seven-component immunological latent state vector (TNF-α, IL-1β, IL-6, IL-10, C3a, NeutrophilActivation, CoagulationActivation) with biologically constrained Jacobian sign enforcement; and (3) the Inflammatory Storm Index (ISI), an acceleration-sensitive temporal integrator producing a continuous 0–1 risk estimate for impending inflammatory storm onset.The framework was evaluated on 803 held-out recordings across LPS endotoxemia, sterile SIRS, and CAR-T cytokine release syndrome datasets, achieving 91.4% classification accuracy, AUROC 0.963, 47.3 minutes mean advance warning time, and 3.2% false positive rate.This repository contains the official open-source implementation of the Vagus-Decipher AI framework, including neural decoding modules, signal processing pipelines, inference architectures, validation utilities, and reproducibility materials.The manuscript is archived as a preprint and has not undergone peer review at the time of publication. Peer review submission is planned to the Journal of Neural Engineering & Biomedical AI.DOI: 10.5281/zenodo.20347323OSF Registration DOI: 10.17605/OSF.IO/3CAQ2