Validation of Software for Patient-Specific Real-Time Simulation of Hepatic Radiofrequency Ablation

CT-guided radiofrequency ablation (RFA) is a potentially curative minimally invasive treatment for liver cancer. Local tumor recurrence limits the success of RFA for large or irregular tumors as it is difficult to visualize the tissue destroyed. This study was designed to validate a real-time software-simulated ablation volume for intraprocedural guidance.

Software that simulated RFA physics calculated ablation volumes in 17 agar-albumin phantoms (7 with a simulated vessel) and in six in-vivo (porcine) ablations. The software-modeled volumes were compared with the actual ablations (physical lesion in agar, contrast CT in the porcine model) and to the volume predicted by the manufacturer's charts. Error was defined as the distance from evenly distributed points on the segmented true ablation volume surfaces to the closest points on the corresponding computer-generated model, and for the porcine model, to the manufacturer-specified ablation volume.

The average maximum error of the simulation was 2.8 mm (range to 4.9 mm) in the phantoms. The heat-sink effect from the simulated vessel was well-modeled by the simulation. In the porcine model, the average maximum error of the simulation was 5.2 mm (range to 8.1 mm) vs 7.8 mm (range to 10.0mm) for the manufacturer's model (p = 0.009).

A real-time computer-generated RFA model incorporated tine position, energy deposited, and large vessel proximity to predict the ablation volume in agar phantoms with less than 3mm maximum error. Although the in-vivo model had slightly higher maximum error, the software better predicted the achieved ablation volume compared to the manufacturer's ablation maps.