Risk prediction of pleural effusion in lung malignancy patients treated with CT-guided percutaneous microwave ablation: a nomogram and artificial neural network model

Abstract

Objectives To develop an effective nomogram and artificial neural network (ANN) model for predicting pleural effusion after percutaneous microwave ablation (MWA) in lung malignancy (LM) patients. Methods LM patients treated with MWA were randomly allocated to either the training cohort or the validation cohort (7:3). The predictors of pleural effusion identified by univariable and multivariable analyses in the training cohort were used to develop a nomogram and ANN model. The C-statistic was used to evaluate the predictive accuracy in both the training and validation cohorts. Results A total of 496 patients (training cohort: n = 357; validation cohort: n = 139) were enrolled in this study. The predictors selected into the nomogram for pleural effusion included the maximum power (hazard ratio [HR], 1.060; 95% confidence interval [CI], 1.022–1.100, p = 0.002), the number of pleural punctures (HR, 2.280; 95% CI, 1.103–4.722; p = 0.026) and the minimum distance from needle to pleura (HR, 0.840; 95% CI, 0.775–0.899; p < 0.001). The C-statistic showed good predictive performance in both cohorts, with a C-statistic of 0.866 (95% CI, 0.787–0.945) internally and 0.782 (95% CI, 0.644–0.920) externally (training cohort and validation cohort, respectively). The optimal cutoff value for the risk of pleural effusion was 0.16. Conclusions Maximum power, number of pleural punctures and minimum distance from needle to pleura were predictors of pleural effusion after MWA in LM patients. The nomogram and ANN model could effectively predict the risk of pleural effusion after MWA. Patients showing a high risk (>0.16) on the nomogram should be monitored for pleural effusion.