A neural network estimation model based light dose control method and system for low-temperature photothermal therapy

Abstract

Photothermal therapy (PTT) is a new tumor treatment method, which has the advantages of high biosafety and non-radiation. Due to the lack of a non-invasive temperature detection method, it is hard to monitor the treatment temperature during therapy, which eventually leads to the fact that the therapeutic light dose cannot be closed-loop regulated. To achieve accurate light dose control to improve the safety of PTT, we propose a neural network estimation model based light dose control method and system for low-temperature photothermal therapy, which can automatically obtain the target temperature by using the neural network trained with photoacoustic(PA) information and accurately adjust the therapeutic optical dose through fuzzy logic in real-time. An in vitro tissue experiment based on pig liver is used to verify this method. The RMSE of the temperature measurement and temperature control of the proposed method are 0.44 and 0.59℃, and the adjustment time and control overshot are 56 s and 0.2%, showing that this method has good dynamic and static characteristics. The anti-interference experiment verifies that this method has good robustness and security. This method has the potential to enable high-accuracy temperature measurement and control in PTT where intrusive temperature sensors cannot be installed.