Capturing Material Thermal Responses under RF-Induced Hyperthermia: An Experimental Imaging Database

Background thermography is widely used to quantify surface temperature distributions during radiofrequency heating in biomedical, material, and industrial contexts. However, reproducible datasets capturing thermal responses of biological and inert materials under controlled RF excitation are scarce. This limitation hinders the validation of electromagnetic–thermal models, the benchmarking of image-based temperature estimation techniques, and the development of super-resolution and reconstruction algorithms. To address this gap, we generated a curated thermal image dataset that documents how different materials absorb and dissipate RF energy within an ISM band. Methods We constructed a controlled RF heating system consisting of a signal generator, a 6 W power amplifier, a circulator with matched load, and a unidirectional applicator antenna operating near 2.5 GHz. Biological tissues (pork, beef, chicken) and inert materials (foam, wood, phantom) were exposed to continuous-wave RF excitation while surface temperature was monitored using three infrared cameras with low, medium, and high resolution. Thermal images were acquired under systematically varied conditions, including RF power (0–5.5 dBm, extended to 6.5 dBm for low-loss materials), camera-to-sample distance (20–30 cm), and observation angle (0°, 30°, 60°). Each recording was captured after stabilization periods to ensure repeatability, and all images were exported as calibrated per-pixel temperature maps. Metadata documenting material type, acquisition geometry, and temperature range were embedded following a standardized naming structure. A complementary data-augmentation pipeline generated resolution-specific synthetic variations, yielding additional controlled perturbations for algorithm development. Conclusions The dataset comprises 3,600 original and 600 augmented thermal images spanning nine material categories and three spatial resolutions. Its standardized acquisition protocol and comprehensive metadata enable reproducible studies on RF–material interactions, temperature-field reconstruction, geometric correction, and super-resolution enhancement. This resource is expected to support research in computer vision, hyperthermia modelling, and the development of machine learning methods for thermal imaging under controlled RF exposure.