Predictive role of peritumoralradiomics on benign and malignant pulmonary nodules

HUANG Shenyang; SU Xiaoyang; HUANG Jian; LIU Huan; HUANG Qianzhun; ZHOU Tuxin; CAI Peikun; HU Botao; HUANG Sumei

doi:10.3969/j.issn.1674-4136.2025.05.011

Abstract：Objective To explore the efficacy of different ranges of peritumoral radiomics modles combined with clinical imaging information on the differentiation of benign and malignant pulmonary nodules.Methods Clinical and CT imaging data from 190 patients with pulmonary nodules in Maoming People's Hospital from February 2018 to July 2021 were retrospectively collected.The dataset was divided into training group and validation group in a 7∶3 ratio.Based on preoperative chest CT images,regions of interest(ROI)of 3 mm,5 mm,and 10 mm peritumoral area were delineated,and the features were extracted,and screened and modeled.Model performance was compared across these areas,and Radiomics scores(Radscore)were calculated.Combined with clinical imaging features,the nomogram of clinical imaging modle,omics model and fusion model was developed respectively,and compared with the MAYO and Peking University models.Additionally,85 patients from August 2021 to August 2022 were collected as the test group,and the modle was externally validated.Results The radiomics model in the range of 3 to 5 mm around the tumor had better predictive performance than other ranges.The AUC values in the training group were 0.939 and 0.936,while those in the validation group were 0.800 and 0.740.In the test group,the AUC values were 0.684 and 0.715.The Radscore of 5 mm peritumoral area had a higher correlation coefficient than other areas in Logistic regression analysis.The fusion model constructed by combining clinical imaging information had better prediction efficiency than the clinical model and radiomics model,with an AUC of 0.956 in the training group and 0.851 in the validation group,and 0.821 in the test group.Conclusions Radiomics features within the 3-5 mm peritumoral range are more effective in differentiating benign from malignant pulmonary nodules.Radiomics models combined with clinical imaging information can enhance the predictive efficacy for preoperative benign and malignant nodules,thereby aiding in clinical decision-making.