Scott T. Clegg

Sensitivity of hyperthermia trial outcomes to temperature and time: Implications for thermal goals of treatment

PURPOSE In previous work we have found that the cumulative minutes of treatment for which 90% of measured intratumoral temperatures (T90) exceeded 39.5 degrees C was highly associated with complete response of superficial tumors. Similarly, the cumulative time for which 50% of intratumoral temperatures (T50) exceeded 41.5 degrees C was highly associated with the presence of > 80% necrosis in soft tissue sarcomas resected after radiotherapy and hyperthermia. In the present work we have calculated the time for isoeffective treatments with T90 = 43 degrees C and T50 = 43 degrees C, respectively, using published thermal isoeffective dose formulae. The purpose of these calculations was to determine the sensitivity of treatment outcome to variations in thermal isoeffective dose. METHODS AND MATERIALS The basis for the calculations were the thermal parameters and treatment outcomes in three patient populations: 44 patients with moderate or high grade soft tissue sarcoma treated preoperatively with hyperthermia and radiation; 105 patients with superficial tumors treated with hyperthermia and radiation, and 59 patients with deep tumors treated with hyperthermia and radiation. RESULTS The thermal dose values calculated are strongly associated with outcome in multivariate logistic regression analysis. Simple dose-response equations result from the analysis, and we use these equations to assess the sensitivity of outcome upon variations in thermal dose. This information, in turn, allows us to estimate the number of patients required in Phase II and III trials of hyperthermia and radiation therapy. CONCLUSIONS For regimens of 5 to 10 hyperthermia treatments, improvements in median T90 (superficial tumors) and T50 (deep tumors) parameters by 1.2-1.5 degrees C could result in response rates high enough (compared to radiotherapy alone) to justify Phase III trials. A similar improvement in response rates would require an increase in overall duration of treatment by a factor of 3 to 5. This would be difficult to achieve while also avoiding thermal tolerance induction. Achieving these temperature goals may be possible with improvements in hyperthermia technology. Alternatively, there may be ways to increase the sensitivity of cells to temperatures that can be achieved currently, such as pH reduction or chemosensitization.

Magnetic Resonance Thermometry During Hyperthermia for Human High-Grade Sarcoma

PURPOSE To determine the feasibility of measuring temperature noninvasively with magnetic resonance imaging during hyperthermia treatment of human tumors. METHODS The proton chemical shift detected using phase-difference magnetic resonance imaging (MRI) was used to measure temperature in phantoms and human tumors during treatment with hyperthermia. Four adult patients having high-grade primary sarcoma tumors of the lower leg received 5 hyperthermia treatments in the MR scanner using an MRI-compatible radiofrequency heating applicator. Prior to each treatment, an average of 3 fiberoptic temperature probes were invasively placed into the tumor (or phantom). Hyperthermia was applied concurrent with MR thermometry. Following completion of the treatment, regions of interest (ROI) were defined on MR phase images at each temperature probe location, in bone marrow, and in gel standards placed outside the heated region. The median phase difference (compared to pretreatment baseline images) was calculated for each ROI. This phase difference was corrected for phase drift observed in standards and bone marrow. The observed phase difference, with and without corrections, was correlated with the fiberoptic temperature measurements. RESULTS The phase difference observed with MRI was found to correlate with temperature. Phantom measurements demonstrated a linear regression coefficient of 4.70 degrees phase difference per degree Celsius, with an R2 = 0.998. After human images with artifact were excluded, the linear regression demonstrated a correlation coefficient of 5.5 degrees phase difference per degree Celsius, with an R2 = 0.84. In both phantom and human treatments, temperature measured via corrected phase difference closely tracked measurements obtained with fiberoptic probes during the hyperthermia treatments. CONCLUSIONS Proton chemical shift imaging with current MRI and hyperthermia technology can be used to monitor and control temperature during treatment of large tumors in the distal lower extremity.

Relationships among tumor temperature, treatment time, and histopathological outcome using preoperative hyperthermia with radiation in soft tissue sarcomas

The lack of an unambiguous thermal dosimetry continues to impede progress in clinical hyperthermia. In an attempt to define better this dosimetry, a model based on the cumulative minutes during which arbitrary percentages of measured tumor temperature points exceeded an index temperature was tested in patients with soft tissue sarcomas treated with preoperative hyperthermia and conventional radiation therapy. Patients received 5000-5040 cGy at 180-200 cGy per fraction. Hyperthermia was delivered 30-60 minutes after radiation therapy and given for 60 minutes. Patients were randomized between one and two hyperthermia treatments per week for a total of five or 10 treatments, respectively. Lesions were excised 4-6 weeks after completion of hyperthermia/radiation therapy. Successful treatment outcome was considered to be the finding of greater than 80% necrosis of the sarcoma upon histopathologic examination of the resected specimen. Forty-five patients were eligible with thermometry data available in 44 patients. An average of 19 interstitial sites were monitored each treatment per tumor. Sixty percent of tumors had a successful histopathologic outcome. Univariate analysis demonstrated that several descriptors of the temperature distribution were strongly related to treatment outcome; more strongly than nonthermometric factors, such as the number of treatments per week, tumor volume and patient age and more strongly than the commonly used temperature descriptors Tmin and Tmax. Descriptors that incorporated both temperature and time were also superior to the more commonly used descriptors Tmin and Tmax. Multivariate stepwise logistic regression analysis revealed that a descriptor of both the hyperthermia treatment time and the frequency distribution of intratumoral temperatures was the strongest predictor of histopathologic outcome and that the best predictive model combined this time/temperature descriptor and one versus two treatment per week grouping. The more conventional temperature descriptor, minimum measured tumor temperature, did not significantly enhance the predictive power of treatment group. Based on these results, we recommend that descriptors based on both the frequency distribution of intratumoral temperatures and hyperthermia treatment time be tested for relationships with treatment outcome in other clinical data bases. Furthermore, we recommend that temperature descriptors that are less sensitive to catheter placement and tumor boundary identification than Tmin and Tmax (such as T90, T50, and T10) be tested prospectively along with other important thermal variables in Phase II trials in further efforts to define a thermal dosimetry for spatially nonuniform temperature distributions.

Pulsatile blood flow effects on temperature distribution and heat transfer in rigid vessels

The effect of blood velocity pulsations on bioheat transfer is studied. A simple model of a straight rigid blood vessel with unsteady periodic flow is considered. A numerical solution that considers the fully coupled Navier-Stokes and energy equations is used for the simulations. The influence of the pulsation rate on the temperature distribution and energy transport is studied for four typical vessel sizes: aorta, large arteries, terminal arterial branches, and arterioles. The results show that: the pulsating axial velocity produces a pulsating temperature distribution; reversal of flow occurs in the aorta and in large vessels, which produces significant time variation in the temperature profile. Change of the pulsation rate yields a change of the energy transport between the vessel wall and fluid for the large vessels. For the thermally important terminal arteries (0.04-1 mm), velocity pulsations have a small influence on temperature distribution and on the energy transport out of the vessels (8 percent for the Womersley number corresponding to a normal heart rate). Given that there is a small difference between the time-averaged unsteady heat flux due to a pulsating blood velocity and an assumed nonpulsating blood velocity, it is reasonable to assume a nonpulsating blood velocity for the purposes of estimating bioheat transfer.

Computational techniques for fast hyperthermia temperature optimization.

Hyperthermia temperature optimization involves arriving at a temperature distribution which minimizes a stated goal function, the goal function having a biological basis in maximizing tumor cell kill while not exceeding normal tissue toxicity. This involves the computationally intensive process of multiple evaluations of the temperature goal function, requiring repeated evaluations of the power deposition and its corresponding temperature distribution. Two computational schemes are proposed to expedite the temperature optimization process: (1) temperature distribution evaluation by superpositioning precomputed distributions, and (2) using representative tissue groups (rather than every point in the domain) to evaluate the goal function. The application of these schemes is illustrated with a typical optimization problem, as applied to symmetric and asymmetric, heterogeneous models. Application of these schemes reduced the optimization time on a DEC Alpha 1000 4/266 (Alpha is a registered trademark of Digital Equipment Corporation.) from several h to min, with little difference in results. The computational schemes, though demonstrated in the context of electromagnetic hyperthermia, are generally applicable to other forms of nonionizing radiation employed in hyperthermia therapy.

Simulation of Focused, Scanned Ultrasonic Heating of Deep-Seated Tumors: The Effect of Blood Perfusion

Ahstmcr-The temperature distributions produced in deep-seated tumors by focused, scanned ultrasound are simulated. Power depositions are calculated using the Rayleigh-Sommerfeld diffraction integral and are then used as inputs to the two-dimensional bio-heat transfer equation to calculate tissue temperatures. The effects of normal tissue and tumor blood perfusions, tumor location and size, transducer scanning pattern, and transducer characteristics (frequency and f-number) are studied parametrically. The results are presented in terns of the range of applied powers that give acceptable heatings for each situation studied. Low frequency (0.5 MHz), small f-number (0.8) transducers are shown to produce acceptable heatings for a range of scanning patterns and blood perfusions.

Verification of a hyperthermia model method using MR thermometry

Simulation of hyperthermia induced power and temperature distributions is becoming generally accepted and finding its way into clinical hyperthermia treatments. Such simulations provide a means for understanding the complete three-dimensional temperature distribution. However, the results of the simulation studies should be regarded with caution since modelling errors will result in differences between the actual and simulated temperature distribution. This study uses a diffusion weighted magnetic resonance (MR) based technique to measure hyperthermia induced temperature distributions in a three-dimensional space in a non-perfused phantom. The measured data are used to verify the accuracy of numerical simulations of the same three-dimensional temperature distributions. The simulation algorithm is a finite element based method that first computes the electromagnetic induced power deposition then the temperature distribution. Two non-perfused phantom studies were performed and qualitatively the MR and simulated distributions agreed for steady-state. However, due to the long MR sampling time (approximately 4 min), poor agreement between the simulations and MR measurements were obtained for thermal transients. Good agreement between the simulations and fibreoptic thermometry measurements were obtained. The fiberoptic measurements differed from the simulations by 0.11 +/- 0.59 degrees C and -0.17 +/- 0.29 degrees C (mean +/- standard deviation for the two studies).

Hyperthermia treatment planning and temperature distribution reconstruction: A case study

While a great deal of effort has been applied toward solving the technical problems associated with modelling clinical hyperthermia treatments, much of that effort has focused on only estimating the power deposition. Little effort has been applied toward using the modelled power depositions (either electromagnetic (EM) or ultrasonic) as inputs to estimate the hyperthermia induced three-dimensional temperature distributions. This paper presents a case report of a patient treated with hyperthermia at the Duke University Medical Center where numerical modelling of the EM power deposition was used to prospectively plan the treatment. Additionally, the modelled power was used as input to retrospectively reconstruct the transient three-dimensional temperature distribution. The modelled power deposition indicated the existence of an undesirable region of high power in the normal tissue. Based upon this result, amplitudes and phases for driving the hyperthermia applicator were determined that eliminated the region of high power and subsequent measurements confirmed this. The steady-state and transient three-dimensional temperature distributions were reconstructed for four out of the seven treatments. The reconstructed steady-state temperatures agreed with the measured temperatures; root-mean-square error ranged from 0.45 to 1.21 degrees C. The transient three-dimensional tumour temperature was estimated assuming that the perfusion was constant throughout the treatment. Using the computed three-dimensional transient temperature distribution, the hyperthermia thermal dose was computed. The equivalent minutes at 43 degrees C achieved by 50% (T50Eq43) of the tumour volume was computed from the measured data and the three-dimensional reconstructed distribution yielding T50Eq43 = 40.6 and 19.8 min respectively.

Simulation of electromagnetically induced hyperthermia: a finite element gridding method

A finite element gridding method for simulating electromagnetically (EM) induced hyperthermia is presented. The method uses patient CT data as its primary input, with critical structures manually outlined (on a graphics workstation) for explicit demarcation. The paper outlines the various stages involved in mesh creation, including procedures for conforming the finite element representation of critical structures to their smooth boundaries, modelling of heating equipment, and modelling of the outer boundaries. The procedure for generating the finite element model is illustrated for an example treatment. Additionally, the results of computing the SAR in six patients are compared to measured values. The comparison reveals agreement between the model prediction and actual treatment within the limits of measurement error.

Towards the estimation of three-dimensional temperature fields from noisy temperature measurements during hyperthermia

The temperatures at most locations are unknown during clinical hyperthermia because temperature data are obtained at only a few discrete locations. In an attempt to further develop a technique for possibly solving this problem, the feasibility of using state and parameter estimation methods to predict three-dimensional temperature fields during hyperthermia treatments is investigated. Previous studies attempting to solve this problem have been limited to only one or two spatial dimensions. This paper investigates some conditions for which an estimation method can predict the complete three-dimensional temperature field for controlled numerical experiments with additive measurement noise. For the range of perfusion patterns considered, results show that the steady-state temperature field can be estimated to within 1 degree C if there is no measurement noise, no model mismatch, and as few as three measurement locations for seven perfusion zones. The addition of measurement noise degrades the performance of this estimation algorithm, especially when the number of measurement locations is small. Use of Tikhonov regularization of order zero significantly improves the performance of the algorithm for these cases. It was found that there is an optimal regularization parameter which maximizes the algorithm performance. This optimal value is a function of the perfusion magnitude and pattern. The present numerical results indicate that the approach used to solve this difficult and ill-posed problem could potentially be extended to estimate the complete temperature field in more realistic clinical conditions, but considerably more progress must be made before that goal can be reached.