Inverse and transient thermal analysis for rapid hyperthermia therapy planning, delivery and evaluation

Abstract

The efficacy of hyperthermia therapy can be enhanced if a thermal management system is available for therapy planning, delivery and evaluation. The integrated thermal management system is not yet available, though some components of the system have been developed. For example, MIT Hyperthermia Program has developed algorithms for fast forward temperature computation, which include hyperthermia thermal model using Finite Basis Element Method (FBEM) and power model for ultrasound applicators. These components can provide simulated prediction prior to hyperthermia therapy and process evaluation after the therapy. This thesis describes the development of other critical components for the thermal management system: the inverse thermal analysis and the transient thermal analysis. For the inverse thermal analysis, iterative algorithms are used for both the Finite Basis Element Method (FBEM) and Finite Element Method (FEM) to predict the desired power field if the optimal temperature field is given. The simulation results show that both FBEM and FEM predict the optimal power deposition field accurately. FBEM is faster than FEM by an order of magnitude for moderate root mean square (RMS) errors. For the combined inverse thermal analysis that links the optimal temperature field to the control parameters of the energy delivery machine, an inverse algorithm based on source superposition has been developed. Numerical simulations with normalized source array for three simple geometry tumor models have been demonstrated. The simulation results show that the inverse procedure can estimate the optimal control magnitude of each individual source to achieve the optimal temperature field with less than 1⁰C of RMS error.

(cont.) For the transient thermal analysis, a fast algorithm based on source superposition, Green's function solution and Laplace transform has been developed.Various practical transient elements have been formulated. The method is validated by the comparisons to the exact solutions of problems with simple geometries. The validation results show that the numerical results approach the exact solutions as the size of the element decreases. The speed-accuracy comparisons show that the computation time per node is about 0.1 second with temperature error around 0.1 ⁰C, which makes the algorithm very attractive for real-time temperature reconstruction.