Thermal Dose Images Enhance the Prediction of Local Tumor Progression After Multimode Ablation

Abstract

The multimode ablation which uses Radiofrequency heating after a pre-freezing process to treat the tumor, has shown significantly improved therapeutic effects and enhanced antitumor immune response. Unlike surgery, the ablated lesion remains in the body after treatment. Effective treatment evaluation is important to assess the immediate outcome and to monitor disease status over time. We proposed a novel postoperative assessment method combined radiomics features and thermal dose information. Patients with hepatocellular carcinoma and colorectal liver metastases from October 2016 to February 2019 were retrospectively studied. Based on the recorded parameters, intraoperative thermal dose distribution were constructed by a two-step "simulation-fusion" method, taking into account the superposition of energy from multiple treatments. Radiomic features of the tumor, such as grayscale, geometry and texture, were collected from preoperative and postoperative MRI, intraoperative CT and the thermal dose maps. The random survival forest model with input of clinical records, preoperative tumor features and intraoperative thermal dose features was the most competitive (C-index=0.92±0.012, iAUC=0.889, iBS=0.041). The grouping based on risk scores allows quick identification of the levels of long term risk after surgery. Based on the predicted survival probability function, clinicians can predictably enhance monitoring and perform other adjuvant therapies.