A finite difference and machine learning framework for non-Fourier multilayer bioheat transfer under laser irradiation

Purpose This study aims to investigate non-Fourier heat transfer in multilayer biological tissues subjected to laser irradiation and to develop a machine learning-based surrogate framework for rapid and accurate prediction of tissue temperature. Design/methodology/approach A three-layer skin model consisting of the epidermis, dermis and subcutaneous tissue is developed by combining wavelength-dependent optical absorption based on the Beer–Lambert law with the Green-Naghdi non-Fourier heat conduction theory. The governing equations are solved numerically using an implicit finite difference method. Artificial Neural Network and XGBoost models are trained to predict the temperature field from the numerical data. Findings The results demonstrate that wavelength-dependent laser absorption and finite-speed heat conduction significantly influence the thermal response of skin tissues. A transition from surface-dominated to subsurface heating is observed, leading to nonlinear thermal safety boundaries. Among the surrogate models, XGBoost achieves excellent prediction accuracy (R2 > 0.99) while considerably reducing computational time compared with the numerical model. Originality/value The study integrates a non-Fourier bioheat transfer model with machine learning-based surrogate prediction to provide an efficient framework for rapid thermal analysis of laser-irradiated multilayer skin tissues, offering potential applications in laser therapy planning, treatment optimization and real-time thermal monitoring.