Application Of Radiomic Features Of Enhanced Energy Spectrum CT Images In Biological Behavior Evaluation Of Peripheral Lung Cancer

Objective The purpose of this study was to predict the expression of VEGF,EGFR,Ki-67 and MVD in peripheral lung cancer based on the radiomic features of CT-enhanced images and evaluate the feasibility and application value of radiomic features in predicting the biological behavior of pulmonary nodule objectively.Methods In this study,48 patients with peripheral lung cancer confirmed by postoperative pathology were analyzed retrospectively and they underwent dual phase enhanced energy spectrum CT scan before operation.The region of interest(ROI)was delineated on the enhanced CT image(arterial phase + venous phase)using ITK-SNAP and analyzed on Artificial Intelligence Kit(GE Healthcare,life science,China).The key features of VEGF,EGFR and Ki-67 were selected by LASSO logistic regression model.The key features of MVD were selected by linear regression model.The expression levels of VEGF,EGFR,Ki-67 and MVD in tissues were detected by immunohistochemistry as reference standards.ROC curve and area under curve(AUC)were used to evaluate the effectiveness of the model in predicting the expression level of immunohistochemistry indicators,and single variable regression line was used to observe the relationship between MVD of immunohistochemistry indicators and radiomic features.Results According to the expression level of EGFR,Ki-67,VEGF,and MVD,2,3,1and 4 key features were selected for modeling in arterial phase,and 2,4,2 and 2 key features were selected for modeling in venous phase.The AUC values of the arterial EGFR,Ki-67,and VEGF predictive models are 0.78(95% confidence interval [CI]: 0.60 to 0.95),0.84(95%[CI]: 0.70 to 0.98),and 0.71(95%[CI]: 0.56 to 0.86),respectively.In the MVD prediction model,the uniformity feature is the best,and its value is the highest in [-1,1] interval.And the AUC values of the EGFR,Ki-67,and VEGF prediction models of venous phase are 0.76(95%[CI]: 0.59-0.93),0.83(95%[CI]: 0.65-1.00),and 0.77(95%[CI]: 0.64-0.91).The two features of the MVD prediction model are common,and their values are slightly higher in the[-1,1] region.Conclusion(1)The expression of EGFR,Ki-67,VEGF and MVD can be predicted based on the radiomic features of CT enhanced images of peripheral lung cancer;(2)Among the built model of radiomic features,the model of predicting Ki-67 expression in arterial phase and venous phase was the most effective;(3)The analysis of peripheral lung cancer lesions by using the radiomic features of CT enhanced images can evaluate the biological behavior of the tumor in vivo,and provide valuable information for the selection of treatment options and prognosis evaluation.