Simulating the effect of fiducial markers on high-intensity focussed ultrasound treatments of the prostate

Simulating the effect of fiducial markers on high-intensity focussed ultrasound treatments of the prostate

grant

High Intensity Focused Ultrasound, Prostate Cancer Treatment, Supercomputing, Numerical Simulation, k-Wave project, PRACE, 1,000,000 Core Hours

Prostate cancer is one of the most common cancers for men in Europe and a leading cause of cancer-related death. For patients with early-stage disease, the cancer is often treated via the insertion of small permanent radioactive seeds (a procedure known as brachytherapy), or using external beam radiation therapy. The latter procedure usually also involves implanting a small number of gold markers into the prostate to verify the position of the gland between treatments. Unfortunately, for 1 in 4 patients undergoing these treatments, their cancer will recur. However, for some of these patients, further treatment using a different therapy technique is still possible (this is called salvage therapy).

In this regard, high-intensity focused ultrasound (HIFU) has shown a lot of promise. HIFU works by sending a focused beam of ultrasound into the prostate using a transrectal ultrasound probe. At the focus of the beam, the acoustic energy causes sufficient heating via acoustic absorption to kill cells in a localised region (and thus treat the tumour), while the surrounding tissue is left unharmed. The use of HIFU as a salvage therapy for prostate cancer has previously been reported in a number of clinical studies. However, the impact of implanted markers and brachytherapy seeds on the efficacy of the treatment is still largely unknown.

The objective of this project is to use numerical simulations to investigate and quantify the effect of implanted markers on prostate HIFU treatments. Specifically, a numerical model of wave propagation and heat transfer will be used to systematically quantify the effects of individual implanted markers, including the shape of the ultrasound beam, the predicted region of treated tissue, and the treatment time. These will be the first large-scale simulations of this kind, as previous acoustic models have been unable to deal with the computational complexity of simulating these effects.

The model (called k-Wave) is based on an efficient k-space pseudospectral discretization of the nonlinear continuum equations. Spatial gradients are computed using the Fourier collocation spectral method (which is based on the FFT), while the existence of an exact solution to the linearized wave equation is exploited to improve the accuracy of computing temporal gradients. Overall, this allows for a significant reduction in both the number of time steps and the number of grid points per wavelength required for accurate simulations. The code is parallelised using MPI, which partitions the 3D domain across multiple nodes and uses a parallel 3D FFT routine. Each time step requires fourteen global 3D FFTs, and a large simulation may have thousands of time steps, consuming up to 100,000 core hours.

Jaroš Jiří, doc. Ing., Ph.D. (UPSY FIT VUT) , research leader
Cox Ben T. (UCL) , team leader
Treeby Bradley E. (ANU) , team leader
Nikl Vojtěch, Ing. (FIT VUT)
Vysocký Ondřej, Ing. (FIT VUT)
Záň Drahoslav, Ing. (UPSY FIT VUT)