Efficacy of 18F-fluorodeoxyglucose PET/CT parameters in differentiating benign and malignant pulmonary nodules after propensity score matching

Ruxi CHANG; Liang LUO; Ruiyan WANG; Weishan ZHANG; Jungang GAO; Wang YUAN; Xiaoyi DUAN

Return

Efficacy of 18F-fluorodeoxyglucose PET/CT parameters in differentiating benign and malignant pulmonary nodules after propensity score matching

VernacularTitle:倾向性得分匹配后18F-氟代脱氧葡萄糖PET/CT参数在鉴别肺结节良恶性效能中的研究
Author: Ruxi CHANG ¹ ; Liang LUO ¹ ; Ruiyan WANG ¹ ; Weishan ZHANG ¹ ; Jungang GAO ¹ ; Wang YUAN ¹ ; Xiaoyi DUAN ¹
Author Information

1. 西安交通大学第一附属医院PET-CT室,陕西西安 710061
Publication Type:Journal Article
Keywords: propensity score matching; positron emission tomography/computed tomography; radiomics; machine learning; solitary pul-monary nodule
From: Journal of Practical Radiology 2025;41(5):762-766
CountryChina
Language:Chinese
Abstract: Objective To include patients with clinical data matched by propensity scores and to explore the value of 18F-fluorode-oxyglucose(18F-FDG)PET/CT metabolic parameters and radiomics in differentiating benign and malignant solitary pulmonary nod-ule(SPN).Methods A total of 54 patients with SPN(27 benign and 27 malignant)were retrospectively selected,all of them under-went 18F-FDG PET/CT scans.Then the metabolic parameters were analyzed,and the metabolic parameters model was established.After delineating the lesion,imaging features were selected through variance and correlation analysis.The logistic regression was used to build the model,and balance accuracy(bACC)was used to compare the performance of the models.The correlation between meta-bolic parameters and radiomics features was analyzed.Results The maximum standardized uptake value(SUVmax),total lesion uptake(TLU),and coefficient of variation(COV)of malignant were higher than those of benign(P＜0.05).SUVmax and COV had positive predictive value for malignant lesions[odds ratio(OR)＞1,P＜0.05].There was no statistical difference between the performance of the metabolic parameters model and the radiomics model(P＞0.05).There was a strong correlation between radiomics features and metabolic parameters.Conclusion After propensity score matching,metabolic parameters and radiomics show no statistical difference in differentiating benign from malignant SPN.