Recursive Multi-step Prediction of Bioimplants Thermal Effect

Abstract

As the implantable devices, like neural prostheses, become more powerful, the temperature increase caused by their operation is becoming a growing concern. For example, it is reported that a temperature increase greater than 1 C could damage the surrounding brain tissue for neural implants. Thus, it is important to model the thermal effect of implantable devices. Previous literatures model the heat transfer from the bioimplant to its surrounding tissue using the famous Pennes bioheat equation, which are then solved by methods like the Finite Difference Time Domain (FTDT). However, these methods have significant computational cost, which makes them not suitable for applications where online thermal effect prediction is required. Moreover, due to the time varying nature of the human body and the complicated relationship between the power dissipation of bioimplant and its temperature increase, a predetermined thermal model exhibits gradual performance degradation through time. In this paper, a real-time multi-step prediction method is proposed to model the thermal effect of bioimplants. Compared with other human body thermal modeling techniques like FDTD and Fourier transform based methods, the method proposed in this paper is much faster and the computational cost is much lower, so it can be implemented online. The developed method can adaptively update the thermal model based on the real time temperature measurement of the tissue having direct contact with the bioimplant, which is easily available by integrating a temperature sensor into the implant, and accurately predict the future tissue temperature under various operation conditions. The effectiveness of the developed method is demonstrated with a battery powered neural prosthesis, which stimulate the neural tissues with Utah Electrode Array (UEA). The thermal effect of UEA is emulated with a COMSOL model, which has been experimentally validated in previous literatures