Prognostic predictors of radical resection of stage I-IIIB non-small cell lung cancer: the role of preoperative CT texture features, conventional imaging features, and clinical features in a retrospectively analyzed

Abstract

Abstract Background To investigate the value of preoperative computed tomography (CT) texture features, routine imaging features, and clinical features in the prognosis of non-small cell lung cancer (NSCLC) after radical resection. Methods Demographic parameters and clinically features were analyzed in 107 patients with stage I-IIIB NSCLC, while 73 of these patients received CT scanning and radiomic characteristics for prognosis assessment. Texture analysis features include histogram, gray size area matrix and gray co-occurrence matrix features. The clinical risk features were identified using univariate and multivariate logistic analyses. By incorporating the radiomics score (Rad-score) and clinical risk features with multivariate cox regression, a combined nomogram was built. The nomogram performance was assessed by its calibration, clinical usefulness and Harrell’s concordance index (C-index). The 5-year OS between the dichotomized subgroups was compared using Kaplan–Meier (KM) analysis and the log-rank test. Results Consisting of 4 selected features, the radiomics signature showed a favorable discriminative performance for prognosis, with an AUC of 0.91 (95% CI: 0.84 ~ 0.97). The nomogram, consisting of the radiomics signature, N stage, and tumor size, showed good calibration. The nomogram also exhibited prognostic ability with a C-index of 0.91 (95% CI, 0.86–0.95) for OS. The decision curve analysis indicated that the nomogram was clinically useful. According to the KM survival curves, the low-risk group had higher 5-year survival rate compared to high-risk. Conclusion The as developed nomogram, combining with preoperative radiomics evidence, N stage, and tumor size, has potential to preoperatively predict the prognosis of NSCLC with a high accuracy and could assist to treatment for the NSCLC patients in the clinic.