CT-Based Radiomics Models May Predict The Early Efficacy Of Microwave Ablation In Malignant Lung Tumors

Objective: To investigate the value of radiomics models based on CT texture features immediately after microwave ablation(MWA)for evaluating the early efficacy and prognosis of malignant lung tumors.Materials and Methods: The imaging and clinical data of 104 patients with malignant lung tumors treated with MWA at our hospital were retrospectively analyzed.All patients were divided into effective treatment and ineffective treatment groups based on the relevant guidelines and consensus.The ratio of random division between the training and validation sets of patients was 7:3.On the CT images immediately after MWA,the tumor and ground-glass opacity(GGO)around the tumor were mapped layer by layer as a region of interest(ROI),and the radiomics parameters were extracted in the two regions separately.The dimensionality reduction of the parameters was performed by the Least Absolute Shrinkage and Selection Operator(LASSO).In the training set,logistic regression was used to construct three radiomics models: the tumor radiomics(T-RO)model,the peritumor radiomics(P-RO)model and the tumor-peritumor radiomics(TP-RO)model.Single-factor and multi-factor analyses were performed to obtain clinical risk factors,and the combined radiomics model(C-RO)was constructed by combining tumor and peritumor radiomics parameters with clinical risk factors.The efficacy of models was evaluated in the validation set.The cut-off value of the C-RO model was calculated from the receiver operating characteristic(ROC)curve as the optimal threshold value,and the patients were finally grouped and then analyzed for progression-free survival(PFS).Results: The analysis showed that the tumor length diameter and the peritumoral GGO minimum diameter(the shortest distance from the tumor boundary to the ablation zone boundary)were statistically different between the effective and ineffective MWA treatment groups for lung malignancies.Two radiomics features were respectively screened from the tumoral and peritumoral ROIs of CT images,and the tumor,peritumor and combined radiomics models constructed from these parameters all showed good performance in predicting early outcomes after MWA for malignant lung tumors in both the training and validation sets,among which C-RO model had the best performance.The AUC of C-RO in the training set and verification set were 0.896 and 0.801,respectively.The calibration curve illustrated that the C-RO model had better predictive consistency,and decision curve analysis(DCA)confirmed that the C-RO model had higher clinical benefit.Survival analysis revealed that the optimal critical value of the C-RO model’s total score was 0.759,and PFS in the low-risk group(score higher than 0.759)was substantially longer than that in the high-risk group(score below than 0.759,P < 0.05).Conclusion: The tumor,peritumor and combined radiomics models based on CT images immediately after MWA could assess the early efficacy and prognosis of MWA in malignant lung tumors.