WANG Lu, LIU Jieke, XU Fuyang, et al
Journal of Clinical Radiology. 2023, 42(12): 1887-1894.
Objective Poorly differentiated invasive non-mucinous pulmonary adenocarcinoma (IPA),based on the novel grading system,was related to poor prognosis,with a high risk of lymph node metastasis and local recurrence.This study aimed to build the quantitative-semantic model of low-dose CT (LDCT) and evaluate its diagnostic performance for distinguishing the poorly differentiated invasive non-mucinous pulmonary adenocarcinoma (IPA) from well/moderately differentiated IPA manifesting as solid or part-solid pulmonary nodules. Methods A total of 259 patients,who underwent preoperative LDCT scan and were pathologically diagnosed with IPA,were included from July 2018 to December 2021.Nodules were assigned to well/moderately differentiated IPA (n=195) and poorly differentiated IPA (n=64) according to the pathological results.The quantitative features,including average diameter,volume and mean attenuation,the semantic features,including nodule type,shape,margin,interface,lobulation,spiculation,pleural indentation,air bronchogram,vacuole sign and vessel convergence sign,and clinical features,including age and gender,were included in our study.The independent risk predictors and predictive model were determined through multivariable logistic regression.The diagnostic performance was assessed by area under curve (AUC) of receiver operating characteristic (ROC) curve,accuracy,sensitivity,and specificity.Delong test was used for comparisons of the AUCs among independent risk predictors and predictive model. Results Diameter (OR=1.195,95%CI:1.072-1.332,P=0.001),mean attenuation (OR=1.009,95%CI:1.006-1.012,P<0.001),vacuole sign (OR=15.610,95%CI:1.679-145.175,P=0.016) and vessel convergence sign (OR=3.134,95%CI:1.031-9.526,P=0.001) wereindependent risk predictorsfor predicting the poorly differentiated IPA.Those features werefurther selected to build the quantitative-semantic model.The AUC of quantitative-semantic model was 0.937 (95% CI:0.900-0.963),which was significantly higher than any other independent risk factors according to Delong test (all P< 0.05).The accuracy,sensitivity and specificity of the quantitative-semantic model was 0.861 (95% CI:0.813-0.910),0.906 (95% CI:0.807-0.965),and 0.846 (95% CI:0.788-0.894),respectively. Conclusion The quantitative-semantic model of LDCT,which could preoperatively predict the poorly differentiated IPA with excellent diagnostic performance in lung cancer screening,mightoffer clinical value for optimizing therapeutic decisions.