An Integrated Electrochemical-Dielectric Framework for Multimodal Cellular Triage:Physics-Informed Edge AI on ESP32-S3

Abstract

Early detection of cellular anomalies remains challenging in resource-limited settings where conventional diagnostic infrastructure is unavailable. This work presents a unified electrochemical-dielectric framework for multimodal (whole blood and buffer) label-free cellular screening that integrates Hydrogen Evolution Reaction (HER) voltammetry with dynamic ohmic drop compensation, Maxwell-Wagner dielectric spectroscopy with global spectral fitting, and machine learning with physics-informed regularization on an edge computing platform. We postulate that metabolic acidosis manifests as shifts in HER onset potential governed by Nernstian thermodynamics with explicit temperature dependence and kinetic/ohmic correction terms, while genomic anomalies, bacteria, and parasites alter complex permittivity through Hanai mixture dynamics with α and β dispersion differentiation. A 1D Convolutional Neural Network (CNN) with dual architecture and physics-informed regularization processes electrochemical signatures on an ESP32-S3 microcontroller, ensuring thermodynamic consistency during inference through explicit loss function formulation. Theoretical analysis indicates that populationlevel measurements (> 10Exp5 cells/mL) can achieve signal-to-noise ratios sufficient for anomaly flagging, though individual cell detection remains infeasible. We discuss implementation requirements, including dynamic iRΩ compensation via highfrequency pulse, external analog front-ends for impedance spectroscopy above 1 MHz, and uncertainty propagation analysis with error budgets for pH and dielectric parameters. This framework targets triage applications rather than definitive diagnosis, offering a portable screening tool for anomalies in multiple biological matrices.