Enhanced magnetic hyperthermia in graphene-magnetite nanohybrids for cancer therapy: Artificial intelligence-driven validation via Nonlinear Autoregressive with Exogenous Inputs-Long Short-Term Memory (NARX-LSTM) forecasting.

Magnetic hyperthermia offers targeted cancer therapy using alternating magnetic fields (AMFs) to heat nanoparticles. This study evaluates graphene-magnetite nanohybrids (GMNHs) across compositions (Fe3O4:graphene ratios 0-100 %) and AMFs (163-982 kHz, 12.7-23.9 mT) to optimize heating efficiency. A hybrid deep learning model Nonlinear Autoregressive with Exogenous Inputs-Long Short-Term Memory (NARX-LSTM) predicted temperature evolution using experimental data. Results identified F75G25 (75 wt% Fe3O4/25 wt% graphene) and F65G35 (65 wt% Fe3O4/35 wt% graphene) nanohybrids achieved therapeutic temperatures (40-45 °C) under 163 kHz/16.4 mT and 518 kHz/23.9 mT respectively, while F65G35 achieved higher SAR (1.39 W/g) under 518 kHz/23.9 mT. High frequencies (982 kHz) reduced efficacy due to mismatched nanoparticle relaxation dynamics. The NARX-LSTM model achieved high accuracy (R2 ≥ 0.997), enabling precise thermal forecasting. This work highlights AI-nanotechnology synergy for optimized, personalized cancer therapy with minimal toxicity.