Comparison of the Adjoint and Influence Coefficient Methods for Solving the Inverse Hyperthermia Problem

Abstract

An adjoint formulation is derived and used to determine the elements in the Jacobian matrix associated with the inverse problem of estimating the blood perfusion and temperature fields during hyperthermia cancer treatments. This method and a previously developed influence coefficient method for obtaining that matrix are comparatively evaluated by solving a set of numerically simulated inverse hyperthermia problems. The adjoint method has the advantage of requiring fewer solutions of the bioheat transfer equation to estimate the Jacobian than does the influence coefficient method when the number of measurement sensors is significantly smaller than the number of unknown parameters. Thus, it could be a preferable method to use in hyperthermia applications where the number of sensors is strictly limited by patient considerations. However, the adjoint method requires that CPU time intensive convolutions be numerically evaluated. Comparisons of the performance of the adjoint formulation and the influence coefficient method show that, first, there is a critical ratio of the number of measurement sensors to the number of unknown parameters at which the CPU time per iteration required to calculate the Jacobian matrix is the same for both methods. The adjoint method is faster than the influence coefficient method only when the value of the ratio is less than that critical value. For the hyperthermia problems investigated in the present study, this only occurs for cases with a very small number of measurement sensors. This presents a potential problem for clinical applications because the fewer measurement sensors used, the less information that can be gathered to correctly solve the inverse problem. Thus, second, when both techniques were utilized to solve several inverse hyperthermia problems it was found that the total CPU time for the adjoint formulation was larger than that for the influence coefficient method for all of the cases which were solved successfully. That is, all inverse solutions which were successful had ratios greater than the critical value. Thus, for practical hyperthermia problems it appears that the influence coefficient method is preferable to the adjoint formulation.