Planning Needle Placement in Image-Guided Radiofrequency Ablation of Hepatic Tumors

Abstract

In hepatic applications, radiofrequency ablation (RFA) produces ablation extents that are limited in size both as a result of local tissue properties as well as constraints in ablation device design and physics. Because RFA is a focal, nonconformal therapeutic modality, proper placement of the device is an important goal in producing successful treatment so that the resulting ablation extents overlap the detectable tumor as well as a suitably defined margin. This dissertation examines novel methods of treatment planning by using image-guided techniques to improve placement accuracy and computational modeling to predict ablation outcomes given suitable placements. A method is presented to search for needle placement that best satisfies a given therapeutic goal using outcomes predicted by finite element models of ablations. This search technique is applied to simulated scenarios requiring single as well as multiple ablations to study effects of nearby heat sinks on optimal placement. A phantom system is then constructed to conduct ablation experiments performed using a tracked RFA device. The phantom ablation results are compared against ablation extents predicted using computational models given the measured positional data from the tracked device. Metrics to quantify the model accuracy are introduced, and the effects of tracking inaccuracies are analyzed. Finally, the sensitivity of predicted ablations to needle placement inaccuracies is studied theoretically. Sensitivity analysis is conducted via a novel method that couples boundary element and finite element methods to obtain multiple simulations efficiently for different needle placements over a static mesh. This method is used with Monte Carlo simulations to generate a spatial map of the likelihood of ablation success given uncertainties in targeting accuracy. Using this technique, strategies to make treatment plans less sensitive to placement errors are studied. The results of this research demonstrate the feasibility of coupling image-guided techniques and computational modeling to produce predictive treatments plans for RFA that are robust to device placement uncertainties.