In this post we report results relative to the use of GPUs for computing the evolution of tissue temperatures during Radio Frequency Ablation (RFA). In RFA electric energy is applied to the tissues with the purpose of rising tissue temperatures and killing a certain volume of tissues, which is supposed to be malignant (RFA is a form of cancer treatment).

The temperature of tissues increases over time as Radio Frequency (RF) energy is applied. RFA physics can be described using the Bioheat equation, to model the heat diffusion in the tissues, coupled with the the Laplace equation, used to model the deposition of electrical energy into the tissues. These two equations are usually solved with the Finite Element Method (FEM) in three dimensions. Tracking the temperature evolution over time requires re-solving these two equations many times. The simulation of a surgical intervention requires solving these equations several hundreds to a few thousands times, and it is computationally very expensive. The simulation of an RFA intervention of 10 minutes has been reported to take several hours of computing time.

We have developed an RFA Physics Library which uses GPUs for acceleration. Based on this, and on several other optimizations, we have been able to reduce the computation time for a typical intervention to tens of seconds. This significant gain will allow conducting simulations in real-time in the operating room, and to use computer models for guiding RFA procedures.

The video at the top of this post shows the temperature evolution for an umbrella-like electrode. All the computations are performed on a GPU using the CUDA programming language.

The video below shows the 70 degrees Celsius surface, as it evolves during an RFA ablation.