Prediction of Microwave Ablation Recurrence in Pulmonary Malignancies Using Preoperative Computed Tomography Radiomics Models.

BACKGROUND:Assessing the early efficacy of microwave ablation (MWA) for pulmonary malignancies is a challenge for interventionalists. However, performing an accurate efficacy assessment at an earlier stage can significantly enhance clinical intervention and improve the patient's prognosis. PURPOSE:This research aimed to create and assess non-invasive diagnostic techniques using pre-operative computed tomography (CT) radiomics models to predict the recurrence of MWA in pulmonary malignancies. MATERIALS AND METHODS:We retrospectively enrolled 116 eligible patients with pulmonary malignancies treated with MWA. we separated the patients into two groups: a recurrence group (n = 28) and a non-recurrence group (n = 88), following the modified Response Evaluation Criteria in Solid Tumors (m-RECIST) criteria. We segmented the preoperative tumor area manually. We expanded outward the tumor boundary 4 times, with a width of 3 mm, using the tumor boundary as the baseline. Five groups of radiomics features were extracted and screened using max-relevance and min-redundancy (mRMR) and least absolute shrinkage and selection operator (LASSO) regression. Weight coefficients of the aforementioned features were used to calculate the Radscore and construct radiomics models for both tumoral and peritumoral areas. The Radscore from the radiomics model was combined with clinical risk factors to construct a combined model. The performance and clinical usefulness of the combined models were assessed through the evaluation of receiver operating characteristic (ROC) curves, the Delong test, calibration curves, and decision curve analysis (DCA) curves. RESULTS:The clinical risk factor for recurrence after MWA was tumor diameter (P < 0.05). Both tumoral and four peritumoral radiomics models exhibited high diagnostic efficacy. Furthermore, the combined 1 (C1)-RO model and the combined 2 (C2)-RO model showed higher efficacy with area under the curve (AUCs) of 0.89 and 0.89 in the training cohort, and 0.93 and 0.94 in the validation cohort, respectively. Both combined models demonstrated excellent predictive accuracy and clinical benefit. CONCLUSION:Preoperative CT radiomics models for both tumoral and peritumoral regions are capable of accurately predicting the recurrence of pulmonary malignancies after MWA. The combination of both models may lead to better performance and may aid in devising more effective preoperative treatment strategies.