Experimental validation of a thermophysical fluid model for use in a hyperthermia treatment planning system

Accurate hyperthermia treatment planning, monitoring, and evaluation of temperatures in and near fluid volumes in the body requires realistic modelling of heat transport within fluids, which is currently not implemented in available treatment planning packages. Aim of this study is to assess the accuracy of a thermophysical fluid model, developed for treatment planning near fluid volumes. A cubic phantom with inner dimensions of (7 cm)(3) was filled with deionised water. The front, back, top and bottom walls of the cube consisted of PVC, the side walls of stainless steel. The left wall was kept at a constant temperature of 25 or 37 degrees C, the right wall at 1, 2, 5, 10, or 15 degrees C higher. Thermal probes mapped the temperature profile in the central vertical plane perpendicular to the cold and hot walls with a spatial resolution of 5-10 mm. The temperature distributions were compared to simulations using a finite volume-based thermophysical fluid model implementing the Boussinesq approximation to the Navier-Stokes equations, developed as an extension to our in-house developed hyperthermia treatment planning suite. The simulations were performed using three meshes at different resolutions. The fluid model predicts the temperature distribution accurately (random and systematic error <0.1 degrees C, at least 95% of absolute errors <0.2 degrees C) for hyperthermic temperature differences (<5 degrees C within the fluid volume). When the temperature differences reach 15 degrees C, the random and systematic errors increase to 0.3 degrees C and 0.1 degrees C, respectively, with absolute errors up to 1.1 degrees C. The thermophysical fluid model predicts temperature distributions in a convective fluid with sufficient accuracy for hyperthermia treatment planning in and near fluid regions. A mesh with a resolution of 0.25 cm combines accurate results with acceptable computation times. (C) 2020 The Authors. Published by Elsevier Ltd.