Elementwise approach for simulating transcranial MRI-guided focused ultrasound thermal ablation

Abstract

This work explored an elementwise approach to model transcranial MRI-guided focused ultrasound (TcMRgFUS) thermal ablation, a noninvasive approach to neurosurgery. Each element of the phased array transducer was simulated individually and could be simultaneously loaded into computer memory, allowing for rapid (∼2.5s) calculation of the pressure field for different phase offsets used for beam steering and aberration correction. We simulated the pressure distribution for 431 sonications in 32 patients, applied the phase and magnitude values used during treatment, and estimated the resulting temperature rise. We systematically varied the relationship between CT (computerized tomography)-derived skull density and the acoustic attenuation and sound speed to obtain the best agreement between the predictions and MR temperature imaging (MRTI). The optimization was validated with simulations of 396 sonications from 40 additional treatments. After optimization, the predicted and measured heating agreed well (R2: 0.74 patients 1-32; 0.71 patients 33-72). The dimensions and obliquity of the heating in the simulated temperature maps were correlated with the MRTI (R2: 0.62, 0.74, respectively), but the measured heating was more spatially diffuse. The energy needed to achieve ablation varied by an order of magnitude (3.3-36.1 kJ). While this elementwise approach required more computation time up front (the combined simulation matrices were approximately 4.6 times higher than a single large simulation), it could be performed in parallel on a computing cluster. It allows for rapid calculation of the three-dimensional heating at the focus for different phase and magnitude values on the array. We also show how this approach can be used to optimize the relationship between CT-derived skull density and acoustic properties. While the relationships found here need further validation in a larger patient population, these results demonstrate the promise of this approach to model TcMRgFUS.