An ultrasound and photoacoustic dual-mode temperature imaging method and system based on EM-UNet deep learning network

Photothermal therapy (PTT) is an emerging cancer treatment technology that leverages the differential tolerance of healthy and cancerous tissues to temperature to eliminate cancer cells through temperature control. However, the lack of precise noninvasive temperature measurement techniques poses risks of undertreatment and overtreatment during PTT. This study proposes a dual modal temperature imaging method based on a U-Net deep learning network, integrating ultrasound (US) and photoacoustic (PA) imaging to enhance the safety and effectiveness of PTT. The method utilizes the U-Net network to extract boundary features from ultrasound images and map them onto photoacoustic images for accurate temperature reconstruction of the target area. The effectiveness of the algorithm was validated through phantom and ex vivo experiments. Experimental results indicate that the system achieves noninvasive temperature measurement accuracies of 0.56 °C for pig liver and 0.78 °C for pig fat. This dual-modal temperature imaging method provides a more precise temperature measurement tool for PTT, potentially significantly improving treatment accuracy and safety.