Predicted SAR/temperature changes induced by phase-amplitude steering are minimally affected by uncertainties in...

Reliability of absolute specific absorption rate (SAR)/temperature levels predicted by treatment planning is strongly affected by tissue parameter uncertainties. Therefore, regular re-optimization to suppress hot spots can accidentally induce new hot spots elsewhere. Adaptive planning methods to avoid this problem re-optimize with respect to the current predicted 3D-distribution. This strategy is robust if reliability of predicted SAR/temperature changes (i.e., increases/decreases) after phase-amplitude adjustments is minimally affected by parameter uncertainties; this work evaluated this robustness.

We validated the basic concept in an inhomogeneous phantom, followed by a patient model. Uncertainties in electrical conductivity, permittivity and perfusion were mimicked by simulations using 100 random parameter samples from normal distributions. Reliability of predicted SAR/temperature increase/decrease after phase-amplitude adjustments was evaluated. Next, correlations between measured and simulated SAR and SAR changes were determined for phase settings evaluated at the treatment start for a treatment series. Finally, practical use in an adaptive workflow was illustrated.

Local SAR/temperature increases/decreases after phase-amplitude adjustments can be predicted accurately. For the phantom, the measured 28.5% SAR decrease was predicted accurately(28.5 ± 0.7%). In the patient model, predicted SAR/temperature changes were typically accurate within a few percent. For the treatment series, correlations between measured and simulated (relative) SAR changes were much better(R²=0.70-0.82) than for absolute SAR levels(R²=0.29). Predictions of steering effects during treatment corresponded qualitatively with measurements/observations.

Predictions of SAR/temperature increases/decreases induced by phase-amplitude steering are hardly affected by tissue parameter uncertainties. On-line adaptive planning based on predicted changes is thus robust to effectively support clinical steering strategies.