Machine Learning-Assisted Electrical Impedance Spectroscopy for Evaluating Bioelectrical Response in Low-Dose Gamma-Irradiated Mice Treated with Mitragyna speciosa Extract

This exploratory study integrated electrical impedance spectroscopy (EIS), Cole-Cole modeling, multivariate analysis, and machine learning to characterize liver and blood bioelectrical profiles in mice subjected to low-dose irradiation using a mixed-radionuclide source assembly and treated with Mitragyna speciosa extract. Thirty-five male BALB/c mice were assigned to seven groups : control, irradiation only, and irradiation with extract doses of 50–150 mg kg-1 body weight. Irradiated mice received daily 40 min exposures for 7 days, corresponding to an estimated cumulative dose of 11.33 mGy. Ex vivo impedance spectra measured from 100 Hz to 100 kHz were summarized using , , , and . Distinct tissue-specific and non-monotonic patterns were observed. AO-100 had the smallest Euclidean distance from the control centroid in liver (1.42) and blood (2.66). Repeated nested cross-validation identified SVM as the highest-performing liver classifier ( balanced accuracy), whereas random forest and kNN achieved identical highest performance in blood . Nested LOOCV and 500-permutation testing supported non-random internal discrimination . SHAP indicated greater contributions from and in liver and from and in blood. Cole-Cole-based EIS provides a promising exploratory framework, although these findings do not independently establish radioprotective efficacy or external generalizability.