The Value of Radiomics Features Based on High-Resolution CT to Predicting if the Lung Sub-centimeter Pure Ground Glass Nodule is Benign or Malignant

Abstract

Abstract Background The study was to develop a radiomics model based on a high-resolution CT (HRCT) scan to noninvasively analyze the benign and malignant sub-centimeter pure ground glass nodule (pGGN). Methods The study included 235 patients with 251 sub-centimeter pGGN (training cohort: n=176; validation cohort: n=75) who underwent preoperative HRCT scans. The volume of interest was manually delineated in the thin-slice lung window, from which 1316 radiomics features were extracted. The least absolute shrinkage and selection operator (LASSO) was used to select the useful radiomics features. The multivariable logistic regression was used to select the clinically important risk factors. The mean CT value model, imaging features model, radiomics model, and combined model were constructed, and the performance was evaluated by receiving operator characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA). A nomogram based on the combined model was developed. Results Gender, mean diameter, shape, margin, and mean CT value were independent clinical risk predictors for predicting the malignancy of sub-centimeter pGGN, and enrolled them to build the clinical predictive model. A total of 39 radiomics features were selected to build the radiomics predictive model. In the validation cohort, the area under the curve (AUC) of the radiomics model (AUC=0.877) and combined model (AUC=0.898) were higher than the mean CT value model (AUC=0.670) and imaging features model (AUC=0.733) (all P<0.05). Conclusion The radiomics model may be useful in predicting the benign and malignant sub-centimeter pGGN before surgery.