An automated software algorithm for optimizing microwave ablation parameters for treatment of liver tumors.

Purpose: To evaluate the performance of a software algorithm developed to streamline microwave liver ablation parameter selection and to compare performance of this algorithm to that of experienced interventional radiologists.

Methods: Patients who underwent microwave ablation for treatment of liver tumors were retrospectively identified. An automated software platform was developed to select the top three 'best fit' combinations of microwave ablation power, time, and vendor for a given tumor to achieve a 5 mm minimal ablative margin (MAM). Generalized linear modeling was used to compare the performance of the software algorithm and experienced interventional radiologists with respect to selected ablation parameters and estimates of total ablative volume (TAV) and MAM. Statistical significance was set at p < 0.05.

Results: 35 patients were identified who underwent single-antenna microwave ablation for liver tumors. Mean estimated TAV was not significantly different between clinical practice (24.96 cm³, 95% CI: 21.18 - 28.75 cm³) and algorithm-derived parameters (23.89 cm³, 95% CI: 20.04 - 27.74 cm³; p > 0.05), indicating agreement in overall treatment approach. However, the algorithm consistently generated ablation parameter combinations with more favorable estimated MAM metrics and significantly lower variability (first algorithm: -5.33 mm, 95% CI -5.40 - -5.26 mm; second algorithm: -5.83 mm, 95% CI -6.01 - -5.65 mm; third algorithm: -6.06 mm, 95% CI -6.30 - -5.83 mm) compared to interventional radiologists (-1.02 mm, 95% CI -2.02 - -0.03 mm).

Conclusion: Streamlining microwave liver ablation parameter selection using an automated software algorithm reduces variability and improves estimated MAM coverage of liver tumors.