Deep Learning–Aided Noninvasive Monitoring of Skin Tissue Temperature Distribution and Blood Perfusion Rate Based on Fractal Conformal Sensors

Skin thermophysical properties are key for health assessment with real-time monitoring enabling early detection of skin-related issues. A polydimethylsiloxane-encapsulated Peano fractal conformal sensor is fabricated by flexible printed circuit technology to accurately measure skin tissue temperature distribution and dermal blood perfusion rate while maintaining full conformity. The second-harmonic method enables precise thermophysical property extraction without requiring high-precision lock-in amplifiers. A circular heat model and multitask learning convolutional neural network (MTLCNN) facilitate rapid thermophysical property prediction, while a thermal impedance network captures temperature distribution during measurements. The sensor provides stable measurements of thermal conductivity and diffusivity with 6.6% uncertainty at a ∼57.3° bending angle. The MTLCNN model achieves a combined correlation coefficient of 0.9054, demonstrating a superior regression performance. Interactions among thermophysical properties, perfusion rate, and temperature distribution support thermal balance in the human body. This approach offers valuable insights for improving noninvasive health monitoring and diagnostic.