The objective of a hyperthermia treatment is to raise all parts of a tumor to therapeutic temperatures. Since instrumentation currently available to the clinic is invasive, temperatures are measured only at a limited number of points. Thus, the temperatures in the majority of a treated tumor are unmeasured and unknown. This lack of knowledge makes it impossible to perform complete therapy evaluations and to perform proper therapy control. This problem will remain inherent in hyperthermia treatments until a method exists which can predict the unknown tumor (and normal tissue) temperatures from the measured temperatures. Simple curve-fitting techniques will not suffice in general since they have no physical basis and will not be able to accurately predict unmeasured temperatures which are higher or lower than those measured. To do this a model which describes the physical processes in the treatment is needed. We propose to utilize state and parameter estimation techniques to estimate the complete temperature field from knowledge of measured temperatures at a few locations. The approach will use optimization methods to minimize the difference between the measured temperatures and temperatures predicted for those same locations by a bio-heat transfer equation model of the heated region. Minimization will be done by adjusting the model's tissue blood perfusion parameters until a solution is reached. Since all other quantities (power deposition, anatomy, thermal properties) will be able to be determined sufficiently well, a priori, blood perfusion is the major unknown to iterate. The specific algorithms developed will be evaluated by testing their accuracy when used on other clinical data and on data obtained from mathematical simulations of hyperthermia treatments. These simulations will be performed using both a standard bio-heat transfer equation model of the tissue being treated as well as more complex simulations which include large blood vessel effects. Sensitivity analyses of the algorithms will be performed to determine the influence of the model parameters on the predicted results, for both steady-state and transient conditions. Comparative evaluations (accuracy and speed of convergence) will be performed for different minimization methods to determine the best approach to this problem. The successful accomplishment of these goals will allow (1) complete temperature fields to be predicted from measured clinical temperatures and (2) complete, systematic treatment evaluations to be performed based on pertinent thermal dose computations.
Showing the most recent 10 out of 28 publications
